approximation par projections et simulations de monte ... · approximation par projections et...

220
Approximation par projections et simulations de Monte-Carlo des ´ equations diff´ erentielles stochastiques etrogrades. Jean-Philippe Lemor To cite this version: Jean-Philippe Lemor. Approximation par projections et simulations de Monte-Carlo des ´ equations diff´ erentielles stochastiques r´ etrogrades.. Math´ ematiques [math]. Ecole Polytech- nique X, 2005. Fran¸cais. <pastel-00001396> HAL Id: pastel-00001396 https://pastel.archives-ouvertes.fr/pastel-00001396 Submitted on 27 Jul 2010 HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destin´ ee au d´ epˆ ot et ` a la diffusion de documents scientifiques de niveau recherche, publi´ es ou non, ´ emanant des ´ etablissements d’enseignement et de recherche fran¸cais ou ´ etrangers, des laboratoires publics ou priv´ es.

Upload: others

Post on 02-Nov-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

Approximation par projections et simulations de

Monte-Carlo des equations differentielles stochastiques

retrogrades.

Jean-Philippe Lemor

To cite this version:

Jean-Philippe Lemor. Approximation par projections et simulations de Monte-Carlo desequations differentielles stochastiques retrogrades.. Mathematiques [math]. Ecole Polytech-nique X, 2005. Francais. <pastel-00001396>

HAL Id: pastel-00001396

https://pastel.archives-ouvertes.fr/pastel-00001396

Submitted on 27 Jul 2010

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.

Page 2: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

"!#%$'&)(*"+-,.#/*10324*&5$607*98:$60<;=(*&)>@?@ACB6+-*

DFE HGI%EJ KLEJMNOPMN QE

RTSVUWYX8)Z\[^]_[a`b`cS@0dSVegf<h

i jkjkl=monqpYr sut@pvmxw j.syl jkl=mz|~t@pvmxwk |t-pYr kYst@pvmwk q| mw1t/|6 sl%vm q|

kst@pvmwk kpT l=|w1t=pv|Yv| 6t=m~ksx6t@p q|l t/l@myxl@sxq|

N:C @H!EN3 -DKL ¡!

¢£EV¤¦¥c "§¨"©ª¬«3O3©­ ¥c !N!®¢ ¯ D "°d±g²"O³¨"©´ ­PD D"!"°cK6KEN¶µ¨"§¬°c ­PD·3!"¸ E¹º!=«O"I=O­°7¯P° ³3E ¯PE!FE6¨"©­"¢O¯°µ" ¥TO3µ° ¸¥7 -¥d³¨°7³ ³3E »"E£V¬¼O³´¦¯ ­PD·3!"

Page 3: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 4: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

-kr

ª ""EJ P!" EJKLEJ '! MNQE! N VD QEJ MN ¤E!_°7­ ¸ ³ QE¬D· ¬KHEJ !I@·¤¦ª¬E ±¶3°7­ ¸ ³ NQE!QE !KH K6\! K6V! EJ C)QED ¶HKLEJ KLEJ MNQEDcEPJE EJ H7FED3!V! c! EQEEJ K6EN !!! E K6¬ 3LDEDEVD ­EJ¬K6QE V¬¬'KL E K63!¬QE ¤EJKL·cd6QEJ7!! D EJ :6K6h!dQEJ7!aD· c·d6KH

M!5 K:FE !LI@·¤¦ª¬E Y¯3¡ E "FE-DD!6DE K6:EJMN!33! 3aD/!D QEEK6 ±Y::! : · E )'K6· EJN:EH!^ELJEJ) K: EJ KLN) QEEK6c!QEd d! c D6EJ EJ K6N 7K6V!D-K6:E4K5!QE5/!KL-E ¯5EQE :E®DD=!-DYE K6"!H# E !P!D MN!"EJ MNV¤ QEEK6 MNP·K6¬ ! 3! $ ­PE¡:KLQE % E DE K6HEJ D£E"!°7­ ¸ ³ ¤& D MNvKL·N!"K6V! QE¬KL K6P!" D£E ! " EK6 DE P'QED Q¯K6¬ E K6MN"!QED E DE# '$® E!PD MV!®D£E D\!:I@E ECV Q!PV !)DND:(QMND E ¬!QE!=K6 QEJ °c)Q!EQE ¡NKL MV<!D /!EP!aD· D /E ! = E K6%!QE/! = EJ $ \/! EJ! dEPJEEJ H KH E ¬!c QEEJKL·c!# E K6¬E Q5!JE ! K6 ¤MNK6 3! 3 · EJ !QE3E:!KL¡QE Q¯3 3 JE4K5-6!6K6 *QQEJD 6 £EJ 566! 5KL ·V!!5!D ·EJ ¡ :DE¡!°7­ ¸ ³x & D MNcDKL¡D H L!QE¡FEKLPQE +uf-,/.10]aS324.gG MNQE'!T N 6VD QE MV- E! KLEJ KLEJ MN5QE¤D $KL ·V!!I=N·¤ ª¬EJ) E MV¤ºQEEKL ·MN $ / E E!%aD· $

Page 5: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 6: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

i .6t/l%sy~gt

!£EJ =)EJKLEJ = cED E!® VD QE D!TFEJYMNQE ¤ ^§ ¸ ­° ¡ @$ ¤ E!I@ ¤ ª¬E@K6·V! \OP)§ ¸ ­° ELE@a7VDFE LQQEDFEJ KEKE D ¬¬E/ KL EJND:!¡6DN3E K6a¬ E KLEJ ´ 'E QEJPEJ'E K 6N!LvQEEJKL·GVE¡ KL h3aDdQE¬EJ!

MI@N ¤ ª¬EJ KHFEJ ! ! a7E4K' O77!!J E o7KLC·EPEJ K cQEEKL ¤3ED ·HE ®EDDED -K6V! - QE K %ELD!®QE$ ¼¡V3! ¡!$<:D MNbV¤ QEEK6DP E D$ ´ E-! QE$ E:E K -5D£EJ6 & D·!§ ¸ ­°D E" -D)5 OKL D£E/ $ ¼EJ: QE3 /QE DE¬DE$< T ¤ P5! ¡·3!$QN-LI%E E£V -D£EJD¬D MV$ ´ L3VK6D£EJQQE D 7D D vQE aD· d L!7cE K :E D£E $ ¬ 7KH EJ JEFEJH3EJ K QEEK6HKL D£EJ :E !QEJP !! !-LPD!/E V ¡"E4K £E QQEJDE EK6 ¡EJ! H /¡!D·EJ ®K6·¤V! a & D·!=§ ¸ ­P°oE!'K6DE 45DK6QEK9)PP!/ P!QE

@S )f<h .@Gb§¬EJD¬EJ!VD QEJ DL!TFEJ7MVEJ QEDFE7KLEJKLEJ D$\I@N·¤ª¬E5K6V!Q¤ºQEEK6DP E D$£E¬$aD· /QE $

Page 7: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 8: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

|r |ol@~pv|r |w1t@

¢FEJP! E!%KLD°7K6KLEVbµ"3%D£E!K6ELQE ¤ DbE®DK6 b:KLEMN))N EdMN'H )MN dE/ KL5DEJD$¶ £EJD)!5ECE D6!6MN7NKL d!! MN¬! \MN E'!KLKLN$ ª · <! dTED ¢K6D $EKLN ¯3D3°7T²¡E<! E£ LK EDD c!QEFE¤EJ E MN37D 7! D ac7 N E :!FEP c MNQEJ! ¬VD QE MV3 ·E!$ ¢) E3ED E D @!=E£JE !g°7­ xECD@»"E£ 5¼EJ7MN5@K6D D QE K6N!E)D)<ED¬!P'! MNTK KL ! £E D·¡ECD"£ED/! ETMN"¬/ N NKL MV¬ E4¤ C + +

MV¶K NED@V¶!QE3E:QEJ "NKL MV ¢gK6D gD QE K6N %EJ¥c ¥7% ¸ EJ¹ !u«PEJKE!! ECEJDD ®!% =D·% ¢%6LD %MN¢£EV¤¦¥c @§ D QEJ!¯PEJ!FE5¨!E¡·­PK6 MN¡¥c D£E!@EJ3EDD !%^EQE"!®¢ I@D !/ MN5!6FE- MN6I%EJ KLEJ MV ¤ ³3D D 6!ª ·JE b¢) E£EJD E KL ¡ !)E !)D·H!QEP EKHEDKHE!)D·)4¤ MN I@DEKLK:! ªI=O¥ a E!KLD"! QE -¢'@ ¤D EJDHE£D: MN "EJ!HEQEJ )VKL MN)" · H £EJ P·! ¸ V JE QE! !-KL¡MN ¡!5 NN ¶EEJ )MN E · !(a MN !EN EE6 E!%_¯3DFEJ$N¥ EJD¢ NOPKEª FEJ!E)QI@VQ¥b$<²¡EJE) QE ·3QE DEN! ¡!¡E£JE QED I@Dv $EJK6NE ' ! ´ H ¬!"³ DNQ¢E¤ ¥7 dDEK6N3!D3MN3EC3=QEJ E \"K6K6DKLN%!'KLE5^EJKL "K6"EJKL P ®"E !¡D·¡ °c)QK6D\´ E 6G<E¶D) EE EKLE DKLK6D ·PEJELQED=Q EE bEKLE E# QI@D#

Page 9: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 10: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

s.v| q| r st@pl=|

AW ,Ch f 0, [afW

! #"$%&')(+*&, #.-'+*/(+01&'2 */-'35436()+7 #++*8*9 1:)"*; #01&'+*/)(< ))-'+* ==

&U ¶h S?>cZ Y`cf-, [email protected],6UQ]BA<fh[ , Z\egS DC ª¬EJ!¡3J 9E ­PD =!¡ E4K5* ¯EJ ¬3VTJ KLNE ± E K5 F *,CU`cSGIH b[ .0hKJ , [ .U), [ fTW SVW ,CSVe` . LD ­P D ·EJ /=K6!

X ­P D ·EJ /! °7­ ¸ ³ °c!¡! D E ¥c !¡# MNQE/ ·EJ!¡!D·' 9M

*,CU`cS F H U`b`\h f#NY[_egU, [^fW ¶S 2ºfW 0, [afW . O+ ­P Q /!¡# E KLE °cDKLK6 %!5 · E L *,CU`cSP8H eQJ , Z\f ¶S ¶S + fW ,CSX&Uh]af O?R

°cDKLK6 %!5 · E¡ M ª KLK6 E 3 KE EJ ¬! 3N E

ST! #" 1-!UV(W%&'X+*Y(+01&2 #'*.-'!Z[(+\ #++*X*9 ]4:)"*9 #0\&+*/)(< ))-'+* ^_

#-S .0h[a`, [afW \S®]¦2 U]`A<f<h[ , Z\egS C) ¯3 ¡!D ·EJ -/KL!¡# °d­ ¸ ³ E

Page 11: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

­PD = E E ±YVDE EJ '!¡# MNQEJ- E! E E ) § !"! "QEJE: VD£E E E E ) °c! QEFEHVDE EJ E MF W¶UQ] .S ¶S]¦24UQ]BAf<h[ ,CZbeS + ¯3 E!:# EQEJ V"! EJ4K6 °c EDPD! ª V6D 3!¡D ¯KL QD·!¡QE1 ° KLN °c ¥cQEE/! I%E EJ -!K6! M ) I%E EJ-!

T1,k,M

M ) I%E EJ-!T2,k,M

M ) I%E EJ-!T3,k,M

M ) I%E EJ-!T4,k,l,M

M ) I%E EJ-!T5,k,l,M

M ) I%E EJ-!P([AM

k ]c) M

+uUf<h U), [afW ¶S]¦24SVh h S\h50feeg[ .S \UW .L]¦24UQ]`A<f<h[ , Z\eS R?R I%E EJ -!¡ E: !PNtk

M M °czN,R

C © E -!3KLEEJ V C ) °c KE -!3K:!¡D C ) dEEJ-KH E !3EEK6 C E ) °c3DKL C

L *'N,CSVW .£[afW . ? ­P V £ OQE ¡!¡# E K6E3 D K6 3 K6 E

+* 1 #+*'* 2=

=[ ,CS ..S \S 0fW SNhKASVW0S ¶UQW .L]aS 0U-. ¶hJ[SNh]_[^WJNUQ[ah S F O ª¬EJ!¡3J P !¡DND

C

Page 12: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

³3 EJDE ­P3 E NK6 !FE E ¥c!/ K6¡´´´ CF ]BA<f<hJ[ ,CZ\egS `cf \h®]^S . J¶U), [^fW. b[JNh SW , [aS]^] S . .,Cf 0Z\U-., [¶S ./hKJ ,Ch f?A<hU \S .hKJ J0Z\[^S . ?O ±Y¡DE!¡NV C ± E K6 C °c!¡VDE EJ C E °c!E-KHFEJ ¬3D·" C °c!¡! D E C M ³3 E

C1,2 ³3FE EJ'!NVP# ED ª D

(+*& #2 #*/1&V(0\&+*

,U`b`cSV] ¶S .50Z¶U` [ ,Ch S .:`bhKJ 0 J\SVW ,/. 9LDRF &)Z¶f[BN ¶S . 2 fTW0, [afTW. ¶S\U-.S O+ «VD V ª ! 7¹ ¥b VK6 QE &feg`\Uh UQ[ .fW ¶S .gU`\`bh f 0Z\S .gU S0f .UW .[_W0hKJVegSVW ,/.bh fLWb[aSVW .U +N¶[FX]^[aUQ[ah S . O?L ¡Uh[^U),C[^fW .£[_e \] ,CUWJNS ¶S.5`\Uh UQeQ@ ,Ch S . ODC

°cKLEJ-!¡# ! E KE 'TE:QE>& E \ 3¡- )E£D¡!T Nv!" E

bPE£DE:QEJ> & bPE£DE:QEJ> & >

bED ! aD·!¡ E \ 3°7­ ¸ ³ & D KEJ

O &feg`\Uh UQ[ .fW \S .5UQ]`A<f<h[ ,CZbeS . ¶S.5`\Uh , [^S .:AHS ,6ACA C)O $:`, [afW .:UeQJNhJ[ 0UQ[^W\S . ?

\ !5DV6D N ) I@ V!!¡ QE EJ

Page 13: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

I@ V!!¡ $E EJ E \ !5DV6D NJEFE M

I@ V!!¡ QE EJ M I@ V!!¡ $E EJ M

±Y 3E£D¡FEHKL ·V!"!±¶ E <¤ ¸ D E ± ¸ M C *'NYSVeg` ]aS . W¶UQW0N[aSNh . R?

E I@V! %^ED "KED ¡!¡ DD M E ª K:QEJ-!¡D£E ECD! N v!¡E-! N · M

E ¨EEJ MV M

E ¨EKL D£E ·3 & D!¡FEH! KL M E W\W\S9NYS . F O k°cMNQEJ!T VD QEMN E! O ° D·3D O ¥c · 3!°7­ ¸ ³ O °cMNQEJ!T VD QEMN E! & D O ´ E DFE MV* MO ³ $ =¤ºQEEKL ·MN MO ¯KH!¡D1 MO ª V6D 3! DFEJK633!K6 ª £O ¬! D D # KLEJ K6 ! DE ! FE(aD·¤ /!¡ $ CO ´ $E !«V ! CO OQE ¡KEJ D £

Page 14: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

w1t/l@m kk~t@pvmxw

-' (+(

ª /E% · 'TD· /!QEJHLD£E!'! DE5ª aD /°cDD L! QED! E!6!QEJ7¬! EK6 ¸ ´¦¯P°c´ ª ¸ ¸ ´ KHEJ :L¯P°c MN\D !# ´ ¤aKEJ /·"ª¬EJD ¸ DQMN 3!QE¡! QEK6"¨ ¸ ´¦³´ ¸ _¨"K6 EJ ¸ ´ KH¤FE o³´ MN% ¸ EJMN :!@FE ! Vo³D D ®·'­P K6N$ ¸ L =FEJ E FE:HE5 DE! HEJª N!HI@EJ KEMN3OPMN 3! °cD¥b ND MN EP¥ca¯3D°c ²¡E\"°cKLKLEN µ"$ »ECV¼gE´ (!6¨ ¸ ´¦³´ ¸ <E6E " VY!¡EHD °7­ ±Y E !D· · E 3!·! EQE V3!K6 V!3VKL MN3!JEJE -! QEJK6 EJ -!¡ FEQQED$ O\D-# '!:KLED H!'# DD b°d­ u6EJKL !%E:EDDE:D N)*!®:!:V! )!' V-!'H HDKL!D $ ¨ ¸ ´¦³´ ¸ ELTPK6@!)! ·! 3!JE EJ "!#%$'& (!! QEJ6!FE¡!D6DKLK6DFEJ · ! 7! L!MN"!! QEJ6!# KHP!N ¬!¡°7­ 3E- MN!EEJ !/D¬!" DD !=DKH E QD QEQ# ³3E P K6N")D)! DNEJ" a PQE K6 ² ¸ M !- :EK6H )aKLE ¸ 5MN / - a ' D· D ! ! !QEN'! E=!% DD /D J %KEJD @! ±¶ENE P!¡# "! !QEP!3D!¡V!D "QEJ " EENDNQE!@ /J!# D D "·3JE6D @DNEJ QQEDE 6MNE) <QE£D E K6KL3! E N 5 !# D· D ! QEJ@DE=FE a!' !QE3¡DNEJ$ QOP#¶ ¡ /JPEK6"QD 3!V!D®!: N)EJK6PKE ¡ MN PQEPFEKLK6) )(D !6# @EJD · ± K6 3DEMN D\! )@! ED QEJ *!)+-,%, !6KLEJ

T!6

K \± NLMNdED LD6DEJ5EJE "!6 D E-K6KL

(ST − K)+

T

GY# :D 66K6KL:! E!MVL

Page 15: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯-3!¡# D D P

TST

! QEJ"E a!K

±Yc ¤ EJDN7!7DEc ^c KED c DMN Hc!# D· D Pa ¬E!EVD %!L! KEJ $ Tª ¦¤ ED NN!a3!T DPECD3KEJD QQED 3DEMN3DK6KL"KED ED3 EJ5! ©3!d!T Dd MN3# 53!3KLE ¬DE!dD^K6 E 3!QE¡# ®! ®¡!5# D· D v¥b" "DKL:!5DDE MN!¡ K63!¡ !:FEHK6 Q K6V! " G5D HQEH K6 ª ¸ ¬ ± ¸ ¬H!: )!'KLV! E g!g*!% E !L!'VD5! ¨ ¤ © VD §¬E '!QEJ § E b$ EE d!EKLNd! MN T·¤º!KLE!!3K6V! c:DK6KL aD·®¤º £EJ ! VD¡! ¨" ¤¦©3VD Y­E §¬°7«¯P "EV¤d!3K6V! # :!L: !¡# E !¬!3VD !3±¶ ·N V°c)Q-D QE¬KL ­P M v¬# !!"KLV!!:D QEJK6N!" $ KL ± NV!D· ' ·!E"!%¬KL¬ JE !DNEJ¬V¤ N°7 ·VE¡KL ·V! ! QQEJ§ FEDH ¸ D !QE dED a!QEJ § ¸ E K6·-!®JEJ D ®' ED· !-# E £E EK6KLN gL !DLQEJ)!E$ b±¶HEJNEJ!!@ ) !6 /@D EJN@a ¡E¤QE£E*a# <!P3JEEJ 3NP MN EKLN3QE & DQEJ !%D!¬ ED YMNVJE !¬KLE D EJ QE£! # !PEKLEJ

T ±YcE$d! ·KL 7¡ 5!# PFED !D!QEFE¡ 6!L!$MN$ E (!" & % + E £E V6!3E"!QEJ

0

TKLEJ !3 MNb! PDV P MN)MNHQE£ P

TT# MN)!)! !¡# EN(!:# ED !¡# OP#/DE v"QECJE

(ST −K)+St

! D! EJD ! ENtN !JE ) 3

V0

EJMNE!P# 5!3 MN < P EJ 3!3P·¬! ED QE!¡# ED·S

EV0 +

φdS = (ST −K)+

_ ∫

φdS & D·QE!6a MN !@E/KLEJ :!L MNdN) MV QELEJ $ 6! ED QEJ)·)!6N ^ N 'QE

φ¡·)QEJ

& D·QE¬!" EJD S

^D#VE) KH EJ '!" EJN¬ED P $ ´ 3E!/ V !3MV v ¡=V0

·3¡EJ φ

MVYEJ aNFEHQE ^E D£E¤ 6!QEC @!P ±YD! 7! DP K6V! EJ 5!# !S

!# E!3!'KLK6¬ ED!DE N E EJ $φ

^DKLK6QE K6! ! ED S

!L! D 3E6 ! ED ·!S

KLD V ±YMN)# = D % ¤ EDNP!P ·= P!P!D¤ V $vD·LK6 V! : !6 6E/QEJE_) / Y# vH!

Page 16: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¨"¯3°c»P° ­°±YO «° ¸ °E6QEPEV MNMN \EL3%!QEJN DEJ # Q´ ¶ EK6dQEdDEJ !FE¡QEJ V 3·JEEC¬!3P E $3MNK6K5 MV-K6 'HFEaK6! E QED ´ JE®!D') ! _ KLE¬@ K6"!/

infφ,V0

R(

V0 +

φdS − (ST −K)+

)

V0 +

∫φdS − (ST −K)+

!D ¡ FE QQEJD MN¶KL# DE VD!L!(ST −K)+

· E L! a V0 +

∫φdS

· R K6! MN)EJVD HEJ% FE QQEJD $ T¨%DK6!%K6 = DNV QED3!E:G°7­ N cc!7KEJD c: dEV c¬ MN Ec!EPN! DE c!K6 V!KL·EN!6K6 K5 ":MN "E=FEJ QQED N "QE£ !¡ -FE: EJ "!

)* #X+.):)"'( :D V< ª L¦V'! QEJHDV)6KLL!LJEJE KLED L DK6 D£E /"K6LQEJ6 MN5QEJ ^E K6N:-QE£ ! = c­=KHEN ·! 6!D !6EEJ !QEdDE$ b©3LK65LE= · D 3!PE D£EJ ¸ !¬'

V0

·¬PEJ $ φ

MNV0 +

φdS ≥ (ST −K)+ p.s.

E N!¬¬ T

!PdKL 7')EKLN ! (ST −K)+

ª !PJEJE E) ·! PQE K6P!E °7² M # ±YPKLMN ) !D·"KLE PN¬- · QE EE' £E '!"! - ¡KEJD %QQEJD \©3 E EJ!6E- D£EJ E/ · 6D! L!QEJ ± M M =!'D D D a5!gKLE K6 6E E =MN

V0 +∫φdS ≥

(ST −K)+

­ E7D cN7 · ! T dDK6KL¬!7D 7K6 ! D£E a QEc K6 ¸ D M =P!3E5D D "!:KL K6

infφ,V0

E(V0 +

φdS − (ST −K)+

)2.

ª K6K6cD¶D ¶D !N\!7FE¬KK6 KEJ D£EJ¶¡FE EQQEJc!¡a !D) HE%QE£P D£E% ¡P)a <DPD P · E "a 7QEd KL ¥cQE cD! E E¡KL

Lp _E£Dp > 2

c! ! )¦¤ D£EJ=!%QE£ Tª D\K6·P!DK6!E5! K6 · ¡!

Page 17: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯KL5!-! 5!QE£JE E-: 6 KL EJN$ ±Y'!5D D @EJ !K6 K5

infφ,V0

E∣∣(ST −K)+ − V0 −

φdS+∣∣p.

¨® EKLND E5JE EJ =QE3 @! !T D T±Y!"D)aN !HFE) QEJ JE:G T!=# D "FE) KL

V0

MN "¬ a 5QEVT

H ^EJ ) ¸ 5MN/ !LEJ H aD! H© v­E" D£E" E QQEJD T

VT − (ST − K)+

·¡!QE" E0 ±Y¡!E6 JEJ3D!=D£E3@DKLEEJN

E(U(VT − (ST −K)+)

) E(U(0)

) <´ D E/g !6!# dE=K66MNQE 6) !=FE/D! ±Y/! !T D!# =EJ! Q\DK6K6¡¡3 K6 EV0

\MNP!E N !H!¡# ¸ vE aD· -! P!-!3FE aDN DKL"E5 · ¡ ! "QEK6 ¡!QE ³°7²¡ # °c)Q)E ¡P!3D D %!6KL K6 3®FEJ'QQED QEJDKLK6V0 +

∫φdS−(ST −K)+

MNQ & ·¬ EJc!FE E ¬! 5!:!7FE"! 3!V !3# KEJc3 !E QEDdVDE Vª ·3 ! E- 5! 6!QE ¸ M c ¡Ea! :!E ¸ D M (a ¸ D Qc"5·V! EJ Tª" !D!EVD MV)D E5 3VD ·P! D # ! d±Y=!Eg! ·K6L% @!N

!QEJ(tk)0≤k≤N−1

MN0 = t0 < t1 < · · · < tN−1 < T

g 5!H EJ!@D QE)E@EJ $ ¶ ! KLV!Q:6 :ED·

SMV ! 7±Y/!LE EJ5!=KEJ ·EJ!KL$ °c' ·V!E MNP'a

T $E !

(ST −K)+

¸ a (!# EtN−1

N! EN!QEEDK6¬E£JETN ! KL QEE'E

VtN−1

EJ MV5QEHK6 E"MN5 )!"JE- EJD S

MN# φtN−1

c©3 aHD6!MNQE :! ·KL $ a T

E!EVT = VtN−1

+φtN−1

ST−StN−1

StN−1

adLE 6E5! N ·N<dK6 Q ± )aKE %!¡!¡!: ^EJ) D ®!:a ¡K6V! QE)Q EJ

(Ft)0≤t≤T ¥b)E! 5^E )MN- L!5'6ED)# 'a!

S!)P! D! K6N% KLMN

VtN−1φtN−1

c! JEFE E £EJ KLEcE E ! FtN−1

ª K6KL !QE K6 K6 # DE7NVT

(ST −K)+

6 K6tN−1

JE !!oEE 6E EJ

(VtN−1, φtN−1

)MN3KL K6 N6# DE

VtN−1+ φtN−1

ST−StN−1

StN−1−

(ST −K)+

V±Y3D 3MNQE!EJMN3 ¬E! 3!QEJDPEVD DK6KLKL! DE$ OP v!tN−1

D D !:KL K6 E

(

V + φST − StN−1

StN−1

− (ST −K)+

)2 _ QEKLc¡JEE E £E !LD£E L E5 FtN−1

¤ K6EJ$ \¥7$\tN−2

\! E bMNV) a ctN−1

JEVtN−1

E YNEE3D d7a

Page 18: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¨"¯3°c»P° ­°±YO «° ¸ °KLE MV 7KL·E%EJ!LKL K6 MNL)E= V!

[tN−1, T ] I K6KL"MN

tN−1

G< YE5! KL @ a " N JE [tN−2, tN−1]MN E

VtN−1

tN−1

bOP \ 7JE%! K6FE=JEJ)!a VtN−2

tN−2

3E5EJ "! N KLNφtN−2

!¡ ¡MN KL K6

E

(

V + φStN−1

− StN−2

StN−2

− VtN−1

)2 _ QEKLd EE)E EJ!'DE EJ6 FtN−2

¤ K6E ±YL!: HE K6NMN !/# E0

! K6(V0, φ0)

V0

E ·5D! DKLK6"EHE!¡#

(< #"1- 1&V(0\& %&' *#2 # -& )*#0\& 01&- )2 #0\& :Dª ! "EJVD ECE¬·V-EJN d¬!JE !¬@KEJD DKL $v E !6E£ 3DK6KLN3K6¡@FEJDNK6 MNK6E5K6 ·¤V! ¨$7°d­ xECE PK6NL @# ED ¥7§ ¸ 3MN3E 5!@ !@ KL_) · _) E"KL ·V!!I@·¤¦ª¬E ±Y" D !"FE)K6 V!"" EN¡G

¨= KHM

E D! · /!S

MN -(Sm)1≤m≤M

¨@D D !5 ! _ <±Y3DN3EE3EJ £E

VtN−1

φtN−1

¨%DK6!®E K6PMN) QED)!PDNPP@E)PMN)# =JEL!D D E KEJ -!VtN−1

φtN−1

!QE@QEJD!! KL Q Q\! E!HJE"Td K6¬KLE V5E575:JE"D D VtN−1

·φtN−1DKLK6%! aD! K6!

StN−1

± QEJD!% 5aD -DJE 6· D "!D5! E! aD " p0,N−1(·)

p1,N−1(·)

¨JE/D D !=! K6¡!E KLEJ !VtN−1

·φtN−1

QE¡ NK6 !FEJ H!DV6DNα0,N−1

α1,N−1

\© a D)DV6D )! ·K6 !EVtN−1

≈α0,N−1.p0,N−1(StN−1

)φtN−1

≈ α1,N−1.p1,N−1(StN−1) (α0,N−1, α1,N−1)

! ·KL DKLK6" KL K6 !¡E:MNQE 1

M

M∑

m=1

(

(SmT −K)+ − α0.p0,N−1(SmtN−1

) − α1.p1,N−1(SmtN−1

)SmT − SmtN−1

SmtN−1

)2

.^)

¨H ^ED KLNb# EQEJ E£D_) 7GJ5EPK6FEJD # ED¬QE ¬K6 ¬M

E D· !:I@·¤¦ª¬E !S

ELK6ED # QED)!PEEPEJ £E FtN−1

¤ºKLEYEY YQEJD¶Dv! \QEJp0,N−1(StN−1

)·p1,N−1(StN−1

)

KE

Page 19: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯ © a

(α0,N−1, α1,N−1)! K6 $-! KL

(α0,N−2, α1,N−2)DKLK6P ¬K6K54¤3!

1

M

M∑

m=1

(

α0,N−1.p0,N−1(SmtN−1

) − α0.p0,N−2(SmtN−2

) − α1.p1,N−2(SmtN−2

)SmtN−1

− SmtN−2

SmtN−2

)2

_ EJQEP!5_) ª KL ·V! '!E ¥d§ ¸ E@£ED ! EJQE ECD ± ¸ E¤ KH! -E=EVD N=EgE'!# QE' @ E DMN¶K63!¡3EKLEJ /!FE:EJ $ ¡! E¦¤·KLE ±¶EH! DN ! -KL ·V!¬VK6 MN¬ MN"!QEJ ¥7§ ¸ D D 'gKK65K65 !L# E@D °c\!E ± ¸ \ ED¡!LI@N·¤¦ª¬E =!

SNKH ¡FE-QE 4 H MN·¤º ¶D6MN 5QE aD K6:D£EJ:!QE ¥d§ ¸ ¬¥bJc§ D E! ¸ VD'cd!3K6EDccE D· I=N·¤¦ª¬E ¡!

SQEJ7!d MVc!D! ED EN E Q3KL ·V!3!3JEJE 5E EC ! KL7!K6V!! ¤-

S ª !QEb c E@!E ¥d§ ¸ EDEQE L!LD D' aD !)QEJE¤K6!¡ E4K5MNYG

"KH"!¡!QEJ!¡QEFEJDKLN!=a N

"KH"!"ED3I@ ¤ ª¬EM

D :!dQEJ!aD· (p0,k,p1,k)

!"D QEJMV!EJ!QEEDK6N(tk)0≤k≤N−1

­E¡ D£EJ!:!:KLED HDK6 ·®!S

¡KLV! !'QE P! KL¤K6N¶ EJ! !=DLMNL 6 ·FE-DL! QE¡# E K6DNN=ELDHDEMNP!) <±¶K6!)E£ # E K6-DN@cDK6K6*^E '!N

M

L# )¬L /KH!*aD·!LQE6# )QcgKK6:K6¶)!L! ·K6 ¡V !DND ­ED·6 \KLK6 !=DLDE!LK6H!6KED 6DK6·v!KLE !P:DK6!¬ & D ! !T QEJEK6 !¬EPK6 V! © abFEK6 V!¡@DK6 "!QE3 ¡DE!KEJD ¡DK6 $v £EJD% ^ED! !5# !H £EJ )!QEDH5E DE" ·®!S

¡ N¡!! DN V $MNY3"D£EJ3 N EJN3!QE "DV!3KLED D MV$ ±Y"D MN!56·# ETD¡ E!:!5# E K6)¡NEK6 ! ^E 6gECD') MVEJ )! N VD EMNH ·EJ!b) D !'¯ °cb²¡EQ· !6 MVMN% ED·PK6 K5 EJ /!%MNMVE!EMNVDEY·JEJE -=KLED ¡ DKL$ °c' · < ^ED DK6K63Ed E!!QE FE"P!" E d# E K6_ ¤ ) dE-DE¬!¬ MNQEJ¬!T VD QE MN E!_°7­ ¸ ³MVvN¬!

Page 20: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

°7­ ¸ ³KHEJ D£E :oKLEJ KE MN QED6DK6K6/LE :/D EJ3!

< /(V:<2 #X-B' #*

³ !°7­ ¸ ³uD P3@D !VDEJ!QE (Y, Z)0≤t≤T

MNY D! "G

YT = ξ,

−dYt = f(t, Yt, Zt)dt− ZtdWt.

_

­ED·"Q/ !E: D£EED !5^) /EHD! 'K6E PaD· !¡FE)ED"!¡FEH

(Xt)0≤t≤T! °7­ ¸ <ª D¶!5G

YT = Φ(X),

−dYt = f(t,Xt, Yt, Zt)dt− ZtdWt,

dXt = b(t,Xt)dt+ σ(t,Xt)dWt.

_ E

¨: !P E D dV!D ¥7¥ M *! °7²¡¥ M E cJ7MN \ ! \ Db# D c!7¶ MNQEJ$ MN\E\KLN@EJ $ ¯3E ¬ D!QED¡D QE =3EEN °7²¡¥ M E #! ^E %ECDH!E D£E 3KLEJ KLEJMN(QQEDQEJ D E£D" K6"! DEJ '! $ ª ! DE!/!3! =DE !"QEC (ST −K)+

Qª K6KL/# EH! !: VMN D !K6K6 !7E )c H! 'aMV 5 DY0

bMN JE"!KEJ!!P EJD · ! d MN !PE\!¬:DE EN EJ

Y0

EJ $ ! K6N"^ E $¬! ED EJ77!¬N3! ED· Sb ¬MN EJ¬!KLED EJ¬ QEC

ξ = (ST−K)+T JOP !7 V ! ED · FE3JE

QQEJ)! a MN E H!QE¡E! H!)VH!H# %·¡EDK6·KLND =MN ¨) !D MN :EYt

E3Eb!H a d!)!$D ¤ DVEK6E \ £E YT = ξ

¨/ ! !6EEE " "KL ¬/ECD5_) ­PL $¶ 5!) )DE!LD 6ga 5EQQED ¶# Y!N JEJFE"DKLK6¡ MN E¡# !S ¶°cQE D $ 6QE)E)! E :ED)!FE ª D E! KEJ KLEJ MVK6)QEJ)

^E 3MV NH ENt

t + dt

E5JEFEJ=!JE 3!®a ¡! ·K6 M

Page 21: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯QEFE)JEFEJ-!

SdSt

QEvMNQE"K6 Eδt

Y!QE3 EJD S

G

dYt︸︷︷︸

=

"!#$ &% '( )*(+,

︷︸︸︷

δtSt

# '#(- $.% '

︷︸︸︷

dSt

+ (Yt − δt)︸ ︷︷ ︸

/+00($1' 2$ 3% )4

rdt.^)

¸ (! No@oKLV!-! S

DK6K6/QEJ6KL =/K6V!§ ED¤ ¸ D dStSt

= µdt+ σdWt

3MN¡# /¡!QE# MNQE ^) =NdYt = [rYt + δt(µ− r)]dt+ δtσdWtD¡MNY=E

Zt = δtσ3=KH EQEJ −1

!−dYt = −[rYt + Zt

µ− r

σ]dt− ZtdWt.

°c-Ef(t, y, z) = −(ry+ z µ−r

σ)'DE EJ MNa 3!'!DNE N(!5EJ ^EJ*_) ¨@@! ! 3!DMN3 MN"QEC ! 63!¡E£ ! °7­ ¸ ³uEJVD

< 8:D\ 7 #+*8- %# #+36+&¸ \MN7 QE£:!c# ! !d!

2EJD

S1 ·S2 KEJ MN7 ! E / a !"D Q"!¡ED MV ¡!! =EJD QE-K6

S1 3¨ KE EJ K6N/MN EV!EQE/ E! MN# 66!D6!E 5DE!6! KLED 5 DKL $ ¸ MN6FE!VQEK6MN"!¡D3!=EDS1 ·

S2 "G(

dS1t

S1t

dS2t

S2t

)

=

(µ1

µ2

)

dt+

(σ1

σ2

)(dW 1

t

dW 2t

)

σi

i = 1, 2

:! @FE 4¤ K6/ -!@FE KLEJD/!=EJ 7ª K6KL/o# E ! ! VMN vPE E£!¡3E:· "JEJ! a 3EMN JEDE=! MN ¡QEC ! <ª D\E! 3E x

vMV YD "=a Y

!¡E EJ x

YMN

YT = ξ,

dYt = [rYt + δt(µ1 − r)]dt+ δtσ1dWt − dCt.

_ M

Page 22: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

°7­ ¸ ³δt

cKLN EJN !P ENt

!QE # EDS1 C cDV:VD7!¬DK:¤KLEJ DE ¸ E@ D'H LLMN# JE% MN) )!DMN5E

YT = ξ ^E6E :/DK6KE !/E® D /QE: a _ M d QE °d­ ¸ ³ !)DE!E D"!'VD!DKLKLE¤ ®KEJP¡ EK6 K6NMN5E ®! K6H!: DEJ ®FEH K64D EN¡! ¡!¡ ¬! °d­ ¸ ³ ¥b !" PKL! DK6 ·!P!-KLED ! $EKLN ! D!MN a !"D D "EJ P MNQEJNPQEC®!" VMN E! D!!" MN¬PQEC®!"# '/EQE¬"EJ ¡EJ QQED 5aD" QE3 @@E %KED DK6 ¬KLE QE3EJ @Y E !E · !¡# EN!=a Q­PE3¡KL!L5ED· $ E6!DD D "!: !6G

YT = ξ,

dYt = [rYt + δt(µ1 − r)]dt+ δtσ1dWt + dφt.

_) C § ® KL HDKLK6DFE ^ED !' ! v \)5"!HH ^EJ# JEJ

[0, T [·

T!¡!

φT = ξ − Y0

Q¨"-E:!D"K6¬MN¡ VDφ KLEE QEJ!

σ1Wt

°c- ·'EJ MN /E)QE¬ !"KL3'K6ENφ = 0

aD£EKEJD DK6 cKEJ7LJE!PKK63K6dMV# E6QE! E)gKLCLDMNLE! E=D! φ _ C" QEP EDK6"H°d­ ¸ ³yE®! _) KLE3®) a ! °7²¡¥ M E PK6NMN @( E FEHJ/! °7­ ¸ ³) _ C Q­P¡=3K6NMN ^) C ¬^E FE:KK6EVD MNFE:K6 K5 EJ -!@ MNMNQE!EJ MVVDE a %a ¬D ! °d²¡¥ M E Y!! K63D¡

< 8 %36+: #'*8- :)"(±Y! ¬E£ 7 $EKLN !¡d!K6 aD·7!KLED MNNQEd ·¤V !QEbFE3K6 V! d§FED¤ ¸ D \HKLED dDK6· ¥ EbK6 bD a 6! D£EJ¶ d: )FE@DNE N! K6NH!'# E !% EJ

R 3E- E

rEMN¶3 KH ¡ ENMN v3ED=DK6! QEJ a "QEJ¡ KL § M vOP v

S :KLV! 5§ ED¤ ¸ D $vFE'!VQE¤K6MN"!¡@a !¡DE=G

dYt = [rYt + δt(µ− r)]dt+ δtσdWt − (R− r)(Yt − δt)−dt

_) /LKL

(Yt− δt)− NKLNENdK6N "!FE¡QEMV °cEN

f(t, y, z) =

−ry− z µ−rσ

+ (R− r)(y− zσ)−

'D EJ"aDMNKL3! D£E ECD!/E-! N ·3! N3 MN JEN(!5 !°7­ ¸ ³

Page 23: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯ < 8&, # #( (+:<& )*±Y/°7­ ¸ ³ /E E/E£DEgK6! EJ !®# @! EN-MNDK6K6:5=!

[0, T ] d¨ @MN=# EN:D ==K6

Y

TMN !® EJN

tb

Yt! !!E%DKLKLEJ

ctH! *a

^D!QE!HDK6KE !) O#-N%aKLP D ¡# 4 MNYE ^E YT = Y

−dYt = f(ct, Yt)dt.

¸ b EJ)DE !'D£EJDEE QQE K6%!DHMNH# !H# EN(Yt)0≤t≤T

EJ ^E ¬°d­ ¸ ³ ¨=3D ­° M Y ­° M E !:D¡$ < 8(+0\&2 #'*8& -((++* V# #++*8*#(+!)+*±Y °7­ ¸ ³ N E K6Nd! dECD MNQEJ E6! QE 7K6 £E !5E'K K6 ^E MN5FE'®! K6H!Lª¬EJD 5 N !-# E!!FE= ! ' MVEJ ! N 6VD EMN' bVFE@ aKH !QVKLE²EJD °c)QEJ D £E3D \E\ K6 ¥7¥ M N ¥ E M NYDKLK63@P "!6 EJ!"! )°7­ ¸ ³uE5/! MVEJ @EJ=! QE

∂u∂t

(t, x) + Ltu(t, x) + f(t, x, u(t, x), (∂uσ)(t, x)) = 0,

u(T, x) = g(x)Lt

3# EJ! Q¶QE

Lt =1

2

d∑

i,j=1

ai,j(t, x)∂2

∂xi∂xj+

d∑

i=1

bi(t, x)∂

∂xi.

°c DD !-D/D QE :MN/E£H !/=°7­ ¸ ³E!/KHK6 D£EJ KLEJ KLEJ MN QQED)HMN H!DQE D KLN KL E! EC!3K6 V!VKL MN6D£ED #

ª :6¡DED !-# EQEJ VH! E K6 a¡!:¡EEN P!: ! °7­ ¸ ³ ¥cK6 V!3! E KE -N! ¦!5 · ¡ 3!QEFEH E \ ! E!V!QE QE T!QEJ QE #¢FEJ)a) QEJ) · E ! Q!" ! °7­ ¸ ³ /QE D ·!FE) E !¬E D¬!Z

E MNE: $FE L2 Q¥b"! Q - D QNC(! QE # ­ED·

Page 24: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

³3° ¸ ©P±V\Oc ¸ ¨"§ °7¯P© ¸DFE6K6E6!EE5QE¡´¬!MNQE3FE:V !DND)! E4¤KLEJ /! ∫ T

0ZtdWt

E3! E K6$ <­PE I v TI%E6 EL ·!D EJ E)DE ! MNQEJ !T N ¶VD QEJ MN ·E!b & D ^°7­ ¸ ³3³" ¯ $EJKLN3D EJ!QE3FE5E¡´´´¬@ ·!# E K6 E=DE!3°d­ ¸ ³³ <°c@ !¡D3 ·! MV$ QEJ5¡=D KLEVKL MVH! EJKLEJ %!QE E Yª :D KLE ^MN ::! D ·EJ %![0, T ]

!L EJ!¡!)!EJ(tk)0≤k≤N

·EJ!%QEEN!FE6D! =K6E ¡!# °7­ ¸ ³Nd EJKLEJ c!Y

Z

! Etk

D !¡QE c!3E KE !% ENtk+1

±Y K6LHMN'D'D KLE® MNLD£EDE KE N"! Q QEP!" ED$Y·"# EH!HQE! DEJ ELK6 ·V!K6C D£EJD$ T¯3!QED)HMV b "QEJ^ED !HDKL!HDK6KLNPDED EJ KLND" ED¡·"DK6KLNPHJ¤QE¡# DK6KL !QE3 EJ4K6 ­E § E M E Y§¬EJ @ N6 D KLE-!:!D · EJ®KL"! ^) QE 5¡ KL¡!5E! VD!6¥b \ª ·LE D6 K6·!55QE¡ FE E ¡DN!

ZP!! ! )!DD:QEJE !L N¤ P!-VD!"¥b ª !QEN$PD KLEHVK6 MN ·PK6 K6N EJ D "D£EJDY! N $E!EJ!¡! KL ·3D¡ 3QEE "!QEJ §¬EJ M E ­P¡D KEJ"VK6 MNJ E K6 KL K6 E PN" £EK6 P : $ vOP !QE ­"I%¥ M # ­PEvI%E'"¥7¡D KLE-NK6 MNH ·¤EJ EJ !H°d­ ¸ ³ b MVEJ !T VD QEJ MN) ¤º ·E!$QE @L a¦¤ oD KL I%¥ M :MN%!LNV'D· 66FE E !7DV6DN V¥ EcK6 QFED! HK6QEJ

ξ! ^) d!FE"aKL

g(XT )

XT

¡ELJEJPKL QE ! :°7­ ¸ YE g

! ·H!HDFEC4 P E )«!PFE! :MNQE K6 ­PH$ PD KE'VK6 MNH D :! ! MVEJ @E /! 3QEJ ¬MNQE £EJ ­E ª M E Tª ·JEDHEVD _) E¬®E K6"KLN KL ¤K6 E QE !D MN !MNQED£EJLE £EJ ©33d# d

YDK6KL 3!QE¬ EJ4K63 · E KEJd¬!J! E a φ

"! f <°c=#

Z QEJ3EQE ­E § ­"I%#§ E!V­P 'I@ KL 'D KE)! E KEJ L!*_ QE 66E KLEJ ! KLK6 QE¡6KLK6:!:JEE ¡E £E¤!H§ <ª Db!)%D KLENK6 MN ^EDK6N3 KL K6 E !NP E¬NFED D <ª !QEJN3EJDP !PDND" V!QE §¬­I%# b­E I%¥ ¸ I@ \)E) ! N) $EKLNFE=DD ^E !'E@ g! -°7­ ¸ ³ K6K6NH E% · 'KLFED E)'KLED E £EJ 65E% ! _) KLE!QE:D£E:

f5 ! !QE5!

Za ! !@ £E K6N ­E §¬¥7 b§¬EJ ®·H¥ E)N)gD KLE%NK6 MNL °d­ ¸ ³1 & D

Page 25: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯!g V _ E¡QE ')!D MNH!MNQEJNQD£E \± ! LH!K6FEJD

XQED QEJ!"I%E 5!KH! EJ QdMVTN d! " !PMNQEQD£E ¤ / KEJ Q©3 a K6 33QEJ !¡E /3D· EJ®K6 K6N ^EJD K6N/ MNQEJ ! EK6KLEJ !VQEJKL MN = ¤!¡# °7­ ¸ ³3³ ­E ­"I@ 7­PFE/I@ 3L=K6 V!/!/ !6 MNQEJ! 3 VD QEMN ¤ E!$TQE )!D MN!MVEN ¤

QDEJ !-# E!H! P! KL Y±YEEQE V¡DK6·KLN" & D!3QEEKL·! K6 V! °c)Q!QE § §D QE!·'\ N'%KL ·V!%! E KEJ o ! ! D EJ gKL)! _ E b©35aD!D · EJ gK6TD· "E"KLEMVN"MN:FE%!:D5E KLEJ %®K6: N5!D QEMN' E)!'!D ·EJ HFE'aK6L! EJD:D! \´ ) E d! D MV¬!'D£EJD!"I%E E£V 6¬EJVD D¬ EJDD! a $EJKLN § °7 b±YHEHEQE VHE=QEEJ !

YKLE QE

Z V°7V 3D KLE)NK6 MNKH3!6D ·¬D !KL·'ED3!!¡FEH!VQEK6MN¡§ FED¤ ¸ D

X

7C"!6 "¦V! EJ K5¡MN ·!3D# 7%¡!! QEJP!FE T !E"D QE *=E5QEJ3! E!MN6G D·PE K6QE 3! D E/=K6!# JE

[0, T ]!5# E !"!

N!QEJ

(tk = kh)0≤k≤N−1

h = T

N

! ¡"QEJ!¡K6!¡ E4K5 # E K6P ·EJ!¡@EEN!¡FEH!QEJ

T

!D QEMN EtkEJVD E ! _) E EN=K6! ! V ^) QEJ

ME D/!gI@ ¤ ª¬E ª K6! KL·!%! K66!DV6D N

α0,k

α1,k

·Ytk ≈ α0,k.p0,k(Xtk)

Ztk ≈ α1,k.p1,k(Xtk)

p0,k

·p1,k

N3!3QE3!%aD· $ ¯-E L! D % ' ! E ^ EJN=EJE ´5·-´´5!DDK6·Y\!)E K5¶ b ! b!QEbD ¯3bE d!QEdD£E! KL LE¡D! 5KL QE ! ^) Ec !3FE%aKLφ(XT )

·NC(!HE)QE ´7 DK6K6VD !!QEJ DE-E)D! LKL QE P aD "!¡E:E D¡!X

¯P $E K63MNW

K6KLN5 !/! KL 1 7­PE5:! E K6$ 6EVD

XQEJ:D KLE ! °c:MV=5J

XN 5!65KH M

E DI@ ¤ ª¬E6E/!QEJ3!! D EJ (tk)0≤k≤N

(XN,m

tk)0≤k≤N,1≤m≤M

QO#EXN,mtk+1

= XN,mtk

+ b(tk, XN,mtk

)h+ σ(tk, XN,mtk

)∆Wmk

∆Wk = Wtk+1

−Wtk

(∆Wm

k )1≤m≤MN3!3 K:FE ¬ ! !QE!

∆Wk

Page 26: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

³3° ¸ ©P±V\Oc ¸ ¨"§ °7¯P© ¸­E3FEHQE ´ Q E4K5¡ FE aK6" EN5G→→→ AWb[ , [aUQ]^[ .U), [ fW ¨ ! Q gD !

RM (Y N,I,I,M,mtN

)1≤m≤MQE

Y N,I,I,M,mtN

=

φ(XN,mtN

)

→→→ A , JNhU, [^fW hKJ ,Ch f A<h U ¶S%SVW ,£Se` . g]¦2 [_W.,CUQW ,tk < T

³EMN @E5D !"# E

tk!5QEd!aD·

p0,k

·p1,k

N¨6! dE cD(αMl,k)0≤l≤1DKLK6" KL K6

(α0, α1)!¡E:MNQE

1

M

M∑

m=1

(Y N,I,I,M,mtk+1

+ hf(tk, XN,mtk

, α0.p0,k(XN,mtk

), α1.p1,k(XN,mtk

))

− α0.p0,k(XN,mtk

) − α1.p1,k(XN,mtk

)∆Wmk

)2.°c)QQ=! Q

(Y N,I,I,M,mtk

)1≤m≤ME

Y N,I,I,M,mtk

= αM0,k.p0,k(XN,mtk

)

¯J¬EJKLEJ !(Ytk , Ztk)

7E (αM0,k.p0,k(X

Ntk

), αM1,k.p1,k(XNtk

)) ¨H KLE¦¤MNbMN !# E

tk (αM0,k, αM1,k)

! QVDK6K67) N ) HE K6c KLNK6 K6NE 3 D "! D¬DK6K6N -EVD D N ) ± E K5 N¤ L!LKLE 5DK6·5!QEFE/QEJ 5´ )QE LI

EJ "!¥c D£E! !=D EMN EJNtk¶DMNd MNLFE= D-! ! D

I!QEE KLEJ

Y N,I,I,Mtk

± EJ K6 D ¡ $EKLN¬! ! aD·¬!PD£EJNC!5E:E"´!D ± EJ K6!¡E:QEJ "´´ N EK6N%aK6" E!:G→→→ AWb[ , [aUQ]^[ .U), [ fW ¨"/

αM0,N .p0,N (·) = φ(·) →→→ A , JNhU, [^fW hKJ ,Ch f A<h U ¶SgSW ,CSVeg` . ] 2 [^W .,CUQW ,

tk < T ¨" KH

(∆Wmk )1≤m≤M

6D 5! !QE6!(∆Wm

k )1≤m≤M ¶¨" ! !

(XN,mtk+1

)1≤m≤Mv6D:!

(XN,mtk+1

)1≤m≤M! Q ¡QE

XN,mtk+1

= XN,mtk

+ b(tk, XN,mtk

)h+ σ(tk, XN,mtk

)∆Wmk .

¨=! Q 3E¬D·αM1,k

DKLK6" "KL K6 α1

!

1

M

M∑

m=1

(αM0,k+1.p0,k+1(X

N,mtk+1

)∆Wm

k

h− α1.p1,k(X

N,mtk

))2

D·αM0,k

DK6K6""K6 K6 ¬α0

!

1

M

M∑

m=1

(αM0,k+1.p0,k+1(X

N,mtk+1

) + hf(tk, XN,mtk

, αM0,k+1.p0,k+1(XN,mtk+1

), αM1,k.p1,k(XN,mtk

))

− α0.p0,k(XN,mtk

))2.

Page 27: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯¯JE KEJ !

(Ytk , Ztk)

(αM0,k.p0,k(XNtk

), αM1,k.p1,k(XNtk

)) \O7)MN# K:¤ K6 EJ'!¡D·PE K6 D ! ¬¡VDE EJ '!¡# °7­ ¸ ³ _) E¬EJMN! aD· !"DEJ <¯3NC!5FEHQE ´´ !¡ D

±Y"QEJEK6!:D"! EJ K6¡ HE"!)KLh = T

N

)D ®!¡E! aD "·:KH:!:E D· I=N·¤¦ª¬E M

Y­E¡FE'QE :´ YEQEJ ¤ KLN¬¦V! 3 ¡KL 3EJ K6 ± ! !h

H L!QE) 6 KLL´ MNd! Yg)EEH!) EJ! QE) QE \MND·)H! ! √

h!E) LDE)gE=D! K6QE

Φ(X)! ^) E:aD· ±YD !-E@E D· !

X:H!:VT¤PE!QEJ!@

f

a J) Q´ ·´ <¥7$Q# 3!E@D !6QEJ5! aD· :LEQEJ V /!QE5 - KL/´ 5MN/D·= )KLE 5QEJFE-K6KL5! !LD·E EN!6!D · EJG¶D 6D£EJ)6 ^E£EJ6MN# TJE ) HDEFE=QEEJ !L# !'EaD· c!3E K6 EJ!! ¬d 3QEdE*ac !¡K6!3 EJ ·EJ!:!5# E K6 v°c)¶EQEJ V¡# "!6E K:FE %!'ECc KLd´ ¡d´ MN<Nd!3QEJ!T 3G K63´ EJ 76 KLFE"K5 DE3 MV

M!L# )Q<E 7MN K63´ EJE K6 MV5MN76 ¡! LE KHEJ " aD· !

M·!KLKLN! aD¬!"QE" ¬!E¬# E K6 V± EVDEJ !"D EJ a DN "!¡# QE DK6KL !QEJ3 EJ4K6 ­EdFEQEJ ´´ dEQE Vc!KLE K6¬MNQ7KL·d!d EJ¬!¬ K6´ "´ )! EK6 E!K6N ¬ E¬!- K6´ 5 a EJNc!c7! !ENd c!7K6K6 ! aD· !QEKLE MNK6 !dKH ª DTK6 !^EJJEdSWeVegS5,CSVeg` .3dQEEK6

NM

3KH"!%aD· !¡QE3EJ !¡ D/! ­E FEQEJ 3´´´ V N¬! EDKL K6 E !D !E QE ´3´´ <¯3K6N ! E!=!QE3 "D QE :¡DK6K6!QE3 "D£EJ3 f d £EJ

(y, z) d!DND!E! D E:5K6c! d!QE K6"´ ¡ ·¡EK6 $ ¨-KL= ·MN"# ¬! E KE

Y ¡E 3! !

h® √

h ¥cT!QEPD QE : E P EJ!¡FEHQE "´´¬E/D£E! °7­ ¸ ³ & D ª DD 6!DLgKH L!LEg EK6E)!6E ¡6DE!D! 7KL QE EJ¤º!T!N·¬!QEN !¬ KEJd

Y·ZVaD!3QEEKL·

NM

! aD!¡QEJ °c)Q !QE¡E-EHVK6 MN Y"¡¡! E! ¡ EJ"V!QEJ¡FEQE ´´· QVK6 MNKLN# D!KL MV3 ^E"JE 3 ¤KH E KLN)HQEEJKL·N

M

·5 'K:'!aDH!'QE ¸ D5HNQE7 Q EJ4K63D QE$ V°cNK6NdMN 7 EJ PVK6 MNKLNN3 K6 MV)D %! !QEFE5E¡´´

Page 28: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

³3° ¸ ©P±V\Oc ¸ ¨"§ °7¯P© ¸¯ $EKLN EN !QEFEQEJ 3´´´c3D£E! EK6 ·¤ D£E ®! KL

1 T¯P® P3 NK6 MNK6P!®E4K6E QEJa!LDb!)FE6QE ´´´ ±Y)K6 "E )P)K6 V!)!) QE EJ 5!QE °7²¡²¡¥ + M E K6N D·)# D 5!LFE= !@°d­¤¸ ³3³·3"D!@3QE ¡K6 V!¡!¡ $FEEJ - !QE §ª °d²¡ # ± E -NK6 MN"!¡D3! !:¡DEED "·3 EJV¤VNK6 MNK6K6NN¬ °c)Q!QE¬PD QE EEQEJ!FE)QE¡!" V KLEJ MN!D QE D !3¬E !DE N EJN!:!'V QQED$ \:! E! :! FEJH 'K6V! '§ FEJD¤ ¸ D H

X¡E ¡D£E!5JEJE ! :KED 6 DMN6ECD65K6V!

^ED·Q DNEJ a ¬EJ! ¥c dJE d L!QE 3DE ! '!T ·¤N! E 5! d·7 QdMN3E"6 £E ¬!f

! ¬QEcc EJ5! ! E- cK6QEJD"¡:!6 ¯"EJ " E K6¡ :D£E¡! EJFEMN-6K6:DK6KLN5 :# E K6'5E:= QEJ¤ !! d)QL DEd! 36EJKL D£E ¬L! KL 3

7K6c# K:¤QED·3!¡FEH! KL -!QE# E K6 ª EJ K6¬dKH N!ECENEdQE7E E6EVD d D !K6K6!QE3E: EJ ¨ ¬! E!'MNDE K6PEVD ! § T QE vEJ'¬'DKLK6D6E!D ·K6_ E¡gEJVD 6FE- !6D·L!D 4¤E®QEP E K6 ·E! v¯3HE K6H!T DHE QEPFELKEJ!75! Q 7E KEJ !D QEJMV3 ENc!!D EJ ¯3cD D dT:EKLEJ @!

(Ytk , Ztk)!QEPPQEJD"DP! "QEP!QE! aD

p0,k

·p1,k

N±YDV6D !33E KLEJ :dDdQEN7 L JEJNdLK6¬KL!KL !cD£E E 3M

E D· I@·¤¦ª¬E ¯JHE K6H! )E! E D£EP\ " K6 KL K6 E! § 7QE¡E=K6 D :! EJ 1^¡!K6V! EJ !

X E E

^E DK6 N°c' ·VFE¡K63ED3! § D P!KH M

E DI@ ¤ ª¬E¡!X

!¡D QEMN3 E7!3!D ·EJ : dE5ENM

ED$ ¯JEVD 3 K KLM

E D¬I@·¤¦ª¬E !D EMNQE !KL$ °c'¡¡# "DK6K6) K6Y

KLEP $EJK6"Z

6EV¤VD ¡ EJ¡E=DE! °7­ ¸ ³uECD@E £EHKE E ¤ ¥ E EJ ! §¬­"I@ · I%¥ ¸ I@ # ECJEJN E D QE!PKL ·V!PMN KL·6! :-V =!/DND@5 EJ K6bMN /!1^EJ :MN3K6FE E"K6KLN=QE¡KEJD EJ £E °c)Q¬- E£/E/ ! '! MNQEQD£EJDEJK6N ! § ¥d "· ­"I% #\DMN :QEJ) ECJEJNEL! D!QE) LD£EJH E £E/ : gD£E! H!=MNQEJNQDEJ KE 'N6!H!QE6D=DE$cKLE H-!·V6!QE5

E

Page 29: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯D£EJ"# E £EL Y°c KL ·V!"!HMNQENQDEJ® VDK6K6¡!PDEED !JK6 V!@ EN!(aD· !E

p0,kp1,k

MN "!D£E D!¡D ! 7 ! * )!5MNQED£EJV¤ KEJ <ª !EN$¡KL ·V! N3QE!=3 QEJ3!aD !6ECD·aK6/·LDDE !/ $ d­ %QEJcKK6/# -!LQELJ¤D£EJ¡¡! Q p0,k

·p1,k

!QE¡# "!¡D! 7¹ # E "!5 KL:D5 :VD£EJ D'MV¬ 5 '!QEJ:'K6 V!'KE )E:!QEJ §¬¥7 ­"I% Y ! ¡£EDg 5KH:!6D !- !QE¡# E K6H¡ D '! a QEJ P´ ­ E"QEJ- !QE¬¬QEp0,k

p1,k!"aDE 3!NP®:MN N EJVD VEaD·! !¬D! # Jª DcQE D K6Nb EJ! !K6) b!cE KE \!H ! EDE£D ^E 7D ! D£E D

ª % ®! E K6 7QE ®'!' ' '!'QEJ'! aD$MNLL·EQE V5D LQE6E ª KLK6/L E£5! ¦! ! MV D !K6K6\6KL ·V!LEE6E=! !@ · L L5DE! EKL 4¤D£EJ%^! f

! !QE!Zd!E ± ¸ ¬ KLP!E K6 3DNE PD·LKL ·V!LE/ 6KLN 5!QE ª¬±¶¥7 # \­L¶# ! 5MNL# )DNEFED'KL'!/E E!/!K6 !

XQE: D !5QE6! aDH

^E ¬=E K6PFEJP!QE!KEJP!" EN E*^ EJDETE3 °cT$¶!QE! K6¡! EJ!6! KL vL)LKL·6!6 £EJ "!QE£JEQ!¡# QED¡!¡E)JEJFEJP! EX

^DKLK6!QEJK6 V!!¡MNQEQD£E ¤ QE"K6 Y·# E/D !/D !6aD¡!5QE b¨ !QE "³ M M Y ° F ¶!EQEJ V3! E K6VD 3!¡DvMV3EQE VD# Y&SN`cSVW ¶UQW ,9> 024S .,=]^U`bh SVe[B@Vh S 2ºf[ . > W¶f-,Ch S 0fW\W\UQ[ ..UQW0S?>¶S .fTW , J ,CU ]^[^S .¶S. cf<hJW¶S . d24SNh hS¶h . SVW 2 fTW0, [afTW¶S ,Cf . ]aS .®`bUh UQeI@3,£hS . ¶S]¦24UQ]BAf<h[ ,CZbeS 0S\[7`cSNhegS , ¶S@]aS. 2ºUQ[ahS Uh[aSNh .£[^e \] ,£UQWJVegSVW ,

*"Wf ,Ch S?>024S .,:]^U%`\hSVe[B@Nh S 2ºf[ .

¶S .fW , \[.09 , J .5S9NY`b]_[ 0[ ,£SVegSVW , \[JVh SVW , .50Z¶fT[ N ¶S 2ºfW 0, [afW . \SbU.S\[¬`cSNh£XegS ,,CSVW , \S@hS ,Ch f SVhLW\eQJNh[¶SVegSVW ,5]aS.HhK@;A]^S . ¶S Uh[aU), [afTW .[_e \] ,£UQWJNS

ª ·EQEJ VP¬QEJE a!QEK6 E 3D£E E¬!'¬ N v!PE£DKLK6 ^E JEJDEEK6T ) D@! G®KLEE3D @!EEJ KH E :D! !=EJDL!6D D!6# E K6H !/6DD5'JE H cO N5MV/D/EQE -: & D'!5QEJEK6N

M

·D !%aD· "!5QEJ5!QE¡ "DKLK6H!QE EJ4K6HTD· D·PE K63QEE ¡3K6 ^ EJ!¡!K6 !QEE& P3E ^EJ DKL ±¶E: N @!=E"G¡QE P MN¡!¡FEHQE*!- :EQEJ VKLEJ KE MNK6\ E4K6$CQE dVK6 MNd¶ EJN b! D£EbE!QEJ! !3 K63 FEJ $

Page 30: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

¯¨\Oc´ ¨"¯ ¸

-B(< 2 ­E¬PDP¬D ! ¬!¬ MNQEJ !T N VD QE MN J¤E! : N JE

[0, T ]HED! KL QE d· ! K6 aD· ! =D3!"JEFE¬! E3!¡!K6

dMVJ QE

X Q­ED QE 3MN

X 3 MNa!" MNQEJ -!T ·¤N PVD QE MV

Xi,t = Xi,0 +

∫ t

0

bi(s,Xs)ds+

∫ t

0

σi(s,Xs)dWs, 1 ≤ i ≤ d_

W

d6K6KLN7 6 E!E!6!!K6q

·dσi

! FE ¤ºK6¬ !¡FEHKLEJ Dσ

!¡E d× q

<¯3a!QED QEJ V6G> v`7f-, Z@ .S

∀i 1 ≤ i ≤ d , $ ! $ bi(t, x)

σi(t, x) $ $ #$

!' +- )+# (t, x) ∈ [0, T ] × Rd ±YMN"!¬NV3 K6 E ¬ DEJ¬¬DV6DN

"K6 @! 3!¡D E ­P KL!DMNPEX

!QE EJ D£E 7QED ¨5)d = q = 13=D !ED·/!¡

St3!¡!VQEK6MN""K6V! "!§FED¤ ¸ D

dStSt

= µdt+ σdWt.

°c%EJ MNQENE aKHP! ´º Ut = log(St)

MNQEN!: FEH!VQEK6MNdUt = (µ− σ2

2)dt+ σdWt.OP \E£DDD !K6V! EJ

SU

·S

QH J') \­E) QE ´3·¡´´ v®E-MN5 EJ" N NJEE"FE :NV) E3! aK6 D PFEHKLEJ Dσσ∗ Ov QEJ ´d´´ gE%E MV# E K65! E KLEJ ! °d­ ¸ ³9 HE£D

UDK6K65JEE ! EJ¬ ¬E£D

S ª" FEEJ EMN 3 N EJK6dJE¬ FEQD£E-!

X

*9 #7 #+*­E)LFE/ QEJHE! ¡)DE D $\JE

CD EJNMN%! ! QEJ'!'QEEK6L!® EJ4K6

N,M aD!®QEJ®D'!VDE EJ @!H °7­ ¸ ³ a QE ´´

C"! !:!

TT!DVLDN

!FE!T X

'!¬DEN!"±YD !ED! 6KL QE 3¬!-! f ±¶EJEJ!

C3D QE ^DMNYE a MNKLK6N3 "D£E ! =

M

Page 31: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´¦¯³¨"­©Pª ´ ¨¡¯

Page 32: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

sl-t@pv| t@ mxl@p q|

Page 33: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 34: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

l=|r pol=| j.syl't@pv|

w jkl@|r pv|l sxvymxl@pt@kr | j mxql Ysl 'mY t=pvmw q| kst@pvmwk kpT l=|w1t=pv|Yv|

6t=m ~ksy6t@p q| l t=l=mxyl%syq|

Page 35: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 36: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

¯ E %!ED QE g E K6E ! MNQE! ) VD QE MN5 E!^°d­ ¸ ³" 7ª ·/E'E®! ' !® E D µ±¼ !5QEJE !QEJ3OQE O !=¥cQE

Page 37: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³

Page 38: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

sxql=|. j mut@ '| | t sxvymxl@pt@kr |

¯D! 3!QEJ3DQEJ " MVEJ / E!" JE5GYt = Φ(X) +

∫ T

t

f(s,Xs, Ys, Zs)ds−∫ T

t

ZsdWs

^´ /

f : [0, T ] × Rd × R × Rq → REJ :MN

XFE-!6# MNQEJ! VD QEJ MN@! KL

dG

Xi,t = Xi,0 +

∫ t

0

bi(s,Xs)ds+

∫ t

0

σi(s,Xs)dWs, 1 ≤ i ≤ d,a´

σi

! Ei¤ KL3!FE¡KEJ D

σ!E

d× q ¯3 ! EJO \!bET\! EJb b°7­ ¸ ³ ¨)E3KLEMNb)DK6QEJE E£D%aO /MNDFE@VN !@D! )!

fgFE=! !QED-

XE! D K6FE5! !QEJDH

ω ­EDQE <%E'MN# NV)) P Q °c Q=LG

> v`7f-, Z@ .S%A -# #

f+ +- , + #

|f(t2, x2, y2, z2) − f(t1, x1, y1, z1)| ≤ Cf (|t2 − t1|1/2 + |x2 − x1| + |y2 − y1| + |z2 − z1|) ' (t1, x1, y1, z1), (t2, x2, y2, z2) ∈ [0, T ]×Rd×R×Rq , sup0≤t≤T |f(t, 0, 0, 0)| <∞ > v`7f-, Z@ .S%A

F + ! $ - $ #%$+-,

Φ

L∞ !' ' $ ! $! $ %$ x1

x2 $ +

|Φ(x1) − Φ(x2)| ≤ C supt∈[0,T ]

|x1t − x2

t |.

, |Φ(0)| <∞ 0 , + $ ! $ ! $ +-$ "& +-, [0, T ]

Page 39: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³± V/) -EJ-MN '' MV' a ! ^´ ­P-bd QE!6D :!5 ¡MNL 5D

(f,Φ(X))D5!6QEEJKL· E!E! a E6O /P¡"J´ 5´ vOP ) KL:O 5TK6¡!5DD !# D¡3 D ! '!^´ /

ª K6K6H" E£ MN :!E" NV!D· <H"¡! a5E4¤KLEJ H!6D (Y, Z)

H! ^´ / O\EN7!! D E K6 JdE ! QMN MN3 EMNYN3 !:= LG $2 #2 #'*/< 8"1%2 #" +*+*/*&''V(+7 #!+*#/[.$0hKJ , [ .U), [afW SVW ,CSVeg` .¨LTPMN LE) Ed!3K6

h = TN

^ P d· MN1c!!QEJ!!D · EJ

tk = kh0 ≤ k ≤ N

QO D!QE!! D E%EJVD Xtk

QEXNtk

=D KE5! °c$ `\`bh fN¶[_eU), [afW \S®]^U 0fW \[ , [^fW ,CSNhe[^W\UQ]aS±¶ED! @K6E

Φ(X) ¡EVD HQE

ΦN(PNtN

)T

ΦN " aD·%! ¤K6 (PN

tk)0≤k≤N

D QEF!PI%EJ :!3!K6 d′ ≥ d

!Ndd

K6DKLE! ED !(XN

tk)0≤k≤N

<°c@! EKL$Q@E ¡ ·¤ K6P! EPJEFE ! E"P!HKLE C HEL!VQEK6MN)K6 D !FE6D! =K6QE T¨%HMVPNtk

Ftk

¤ºKLEPMNE[ΦN(PN

tN)]2 < ∞ § = 3DE KEJ -! ! aK63!/ !¡FEHD! 'KL QE ·K6QED·E5KL ¡QE#

E|Φ(X) − ΦN(PNtN

)|2 ¬ K6¡´ !QE "VD QE D QE J­P K6 ! aD· ΦN $EJKLN ! !QE ED·H JE ©®E@NV® 'MV%K6LMN %D E %°7­ ¸ ³ ! K6!D·

(Y Ntk

)k · ¡DK6K6"aD'±YD

yNk (·) !PNtk

a ´ !QEJVD QE LD QE ª · 3E 3!QEJFE¡QEJ 3´ MNQE D K6!33! EJKLEJ > v`7f-, Z@ .S%A

+ $ ! $ ΦN(·) !' '$ #$ #$

N supN |ΦN(0)| <∞ , E|PN,k0,x

tN− PN,k0,x′

tN|2 + E|PN,k0,x

tk0+1− PN,k0,x′

tk0+1|2 ≤ C|x− x′|2 '$ #$#$

k0

N

´ D(PN,k0,x

tk)k

N FEoD EF!I@E C(PN

tk)k

EEN®!PNtk0

= x ­P $

DKLK6@E L!' ! & !(PN

tk)k' LE NEJ

Page 40: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥7´ ³3° %ª O­³3°« ¡¥d¨« ¸ ° ¸ °c O±\µ¨¡³´ «I@°E!QEJ! !=D=D QE/!@I%E DK6K6-/ / DN%E £E '!=E aKLPNtk

= FNk (Uk, P

Ntk−1

)

(FNk )k

"! aD PK6EJ·(Uk)k

N!PJEE E £EJ !

8hf S 0, [^fW . ¶h ¶S . bU-.S . ¶S 2 fTW0, [afTW. ±¶ED·

L2(Ω,P)! 3EEEJ EJ

U(^EJKL Q

φ = [φ1, · · · , φn]∗^D! DK6KL7D· E £EJ T Pφ(U) J¨)! E¶QEJ Rφ(U) =

U − Pφ(U)# !¡JD

OD QEMN" EJNtk¬EVD ¬D· KLN

YtkZl,tk

Zl,tk

FEl¤ºK6PDK:¤EJN3!

Ztk1 ≤ l ≤ q

5JE¡ cD K6 7QE! aD·p0,k(P

Ntk

)pl,k(P

Ntk

)1 ≤ l ≤ q

¶MN DE@Ep0,k

pl,k

1 ≤ l ≤ q

¡ K6 Q ­EL=FE d!%KLE!K6 Q:L D d aE FE Da NQE6!*aD (pl,k(·)

) QELE £EJ (pl,k(P

Ntk

)) ¶­PD D "!

p0,k

·(pl,k)1≤l≤q

N! ¡!QE¡FE'QEJ :´ ¶±Y¡DVLDN!5D α0,k, α1,k, · · · , αq,k

5DK6KL!HD·5D c¨/MNE|pl,k|2 < ∞

0 ≤ l ≤ q6-E-®! E @MN

E(pl,kp∗l,k)

/ N DMNE# D !3DV6D 3!¡FEHD· Ppl,k

0 ≤ l ≤ q

¥b KL Q$N' D

fk(α0,k, · · · , αq,k)

fk(αk)

f(tk, XNtk, α0,k · p0,k, · · · , αq,k ·

pq,k)

\-'K6 Q Ek(.) = E(.|Ftk)

·∆Wk = Wtk+1

− Wtk

∆Wl,kDKLEN¡QEJDKLE

¨=! Q D aD vk

QE[vk]

∗ = (p0,k∗, p1,k

∗ ∆W1,k√h, · · · , pq,k∗ ∆Wq,k√

h)

¥b HD·x |x| ! EK6 D ! ±¶E!K6)FEJ 7 K6 D4 ¥bg

M·x ∈ RM \g! Q |x|2M = 1

M

∑Mm=1 |xm|2

b¥b5^EK6 5!DV6DN3!¡D·α = (α0, · · · , αq)

/ |α| = max0≤l≤q |αl|^ ! KL !

αl ·! N T¥b3^EK6 !aD·!QE !6% E

tk

|pk|3! Q¶!E:KK6"KLE

¥b KEJ D VKL · MN¬ A ‖A‖ · ‖A‖F cD K67 KLE KHK!FEHJE E !3JEJ3·E6K6¡! Q a! QQEJ ‖A‖2

F =∑

i,j a2i,j

[_e \]^U),C[afW.­E7FE

MKHEJ ! !QEJd!

(PNtk

)0≤k≤N(∆Wk)0≤k≤N−1

Nd ¨=J((PN,m

tk)0≤k≤N )1≤m≤M

((∆Wm

k )0≤k≤N−1)1≤m≤M

±Y3JEJ3!(aD·3!QE ·JE P 6!E D· KH 3N (pml,k = pl(P

N,mtk

))0≤l≤q,0≤k≤N−1,1≤m≤M

M

Page 41: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³ ­PcKEcEJQE !

fk(α0,k, · · · , αq,k)fk(αk)

£"fmk (α0,k, · · · , αq,k)

fmk (αk)

f(tk, XN,mtk

, α0,k · pm0,k, · · · , αq,k · pmq,k)

¨=! Q D3D

vmkE

[vmk ]∗ = (pm0,k∗, pm1,k

∗∆Wm1,k√h, · · · , pmq,k∗

∆Wmq,k√h

)

FEHKLEJDV Mk = 1

M

∑Mm=1 v

mk [vmk ]∗

KLEJDPMl,k = 1

M

∑Mm=1 p

ml,k[p

ml,k]

∗ 0 ≤ l ≤ q

(hfW0U),K¶h S.¥bEFE) E !" EJ4K5V- !¬D MN!" E VQE J EJ JEJNG

­E6FE ´ ! $c =QEJEL6!6 K6 5! D D KLNc! aD· !"JE d (ρNl,k)0≤l≤q,0≤k≤N−1

KL 7QE V¨6E ρNk (PN

tk) = [ρN0,k(P

Ntk

), · · · , ρNq,k(PNtk

)]∗

OPVD !%D)K6 $b! Q H! aD _E £EJ "!LD£EJρNl,k(x) =

ρNl,k(PNtk

)ξ(x/ρNl,k(PNtk

))ρN,ml,k (x) = ρNl,k(P

N,mtk

)ξ(x/ρNl,k(PN,mtk

))

ξ : R 7→ R :

aD· /!¡DEC2b

MNξ(x) = x

|x| ≤ 3/2 |ξ|∞ ≤ 2

|ξ′|∞ ≤ 1

B) #"V¨L ! N J3E K6EVD a´ ¤º´ N¯J3VD !DKH/D KE E!=oKL ^!@ EN

tN = T! EN

t0 = 07E K6:!- ) a EN

i = 1, · · · , I EJ)!/¥cD£EJ! ·5FE% !)K6P!)KL!D£EJ QE PM

E DPKH a EJN ¬E! aD¬D D !K6KLN ± E KE LD QE # EJ¤ K63J (Y N,I,I,M , ZN,I,I,M )

KLN¬MNDE KEJ '! !@!"FE! D EgK6b!KHL! EJ!-¥c D£E! aFE= D-!! I

MN ¡!QE "D QE ¡ ·3!=KH"!"E D· 3I@N·¤¦ª¬E M

<¨@E:K6 FE! !QEJD!E aD¬!"QEJ"¬"QEE! !QE£JENE!EJ ! !=DV VFEJ 7 3DKLK6ND 3 EJKLEJ (Y N,I,I,M , ZN,I,I,M )

→→→ AWb[ , [aUQ]^[ .U), [ fW

± EJ K53 EJ ECD

Y N,i,I,MtN

= ΦN(PNtN

)^ ! !QEJKLK6!

iI Q¥c $FEH

(Ytk , Z1,tk , · · · , Zq,tk)!:= EN

tk3 N VFEH!DV6DN3!¡D·

(αi,I,Ml,k )0≤l≤qQE

Y N,i,I,Mtk

= ρN0,k(αi,I,M0,k · p0,k),

√h ZN,i,I,M

l,tk= ρNl,k(

√h αi,I,Ml,k · pl,k)

ρN0,k

ρNl,k

N¶ vD£EJ¶ NV! vEEE£JE £° MV ! bDK6KL DDV6DN¶¶D£E D v¡ EvY £E E ! !QEN((PN,m

tk)0≤k≤N )1≤m≤M

((∆Wm

k )0≤k≤N−1)1≤m≤M

Page 42: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥7´ ³3° %ª O­³3°« ¡¥d¨« ¸ ° ¸ °c O±\µ¨¡³´ «I@°

→→→ A , JNhU, [^fW hKJ3,£hf?A<h U ¶SuSVW ,CSVeg` . ]¦2 [_W.,CUQW ,tk < T

¨ MV ED 3E KEJ Y N,I,I,Mtk+1

:= ρN0,k+1(αI,I,M0,k+1 · p0,k+1)

3=JY N,I,I,M,mtk+1

=

ρN,m0,k+1(αI,I,M0,k+1 · pm0,k+1)

@ ·JE QEJ -FEm

¤ºK6"KHFEJ →→→ ¥bY# EJ EJ

i = 0!\ EJ v!¥c D£E!£

Y N,0,I,Mtk

= 0ZN,0,I,Mtk

=

0#

α0,I,Ml,k = 0

0 ≤ l ≤ q

→→→ ¥b

i = 1, · · · , I T PDV6D αi,I,Mk = (αi,I,Ml,k )0≤l≤q

" EJ K6NVDKLK6" EJK6 '(α0, · · · , αq)

!¡FEHMVEN 1

M

M∑

m=1

(

Y N,I,I,M,mtk+1

− α0 · pm0,k + hfmk (αi−1,I,Mk ) −

q∑

l=1

αl · pml,k ∆Wml,k

)2

.^´

¸ 3@ K6@!%K6 !LDE D4¤ !-EgL ^ dFE KLEJD=! KL MNL :E)¶DMN E E£D5^E5E 6 MNM

!VNE! -QE KLP!FE !K6PK6K6E ¸ D N@P a ! µ ± M ¶ !6MNMNPEJ 3TD P!QE EO °c^EJ DD - PE E EJE5E! D!D¡3 EJQE V!: ±Y:QEEJKL·:!/DD=!/D6E4K6L5'E:!'K6

hh → 0

QE"! aD HKHH!HKHFEJ M

M → +∞ v± KLEDP!:D QEMN5EE¤K6 7EQE V !QE b bVD QE D QE $JD!EN ! E! · EG! D E/=KL3! °7­ ¸ ³ QD 3!3QE!%aD!QE

L2(Ω,P)JDHK6 MVH EH )E DHKH $ ©3- KLEJ!-

QE DK6KL !QE E4K5! D D ¤ !! D EH!D·K6!FEDKHQEJ ¤%!3 K63´ ·¤ ´ ¤ ´ ) Q¯ $EKLNMV ¶P)6EP! TD·I = 3 EJ ¬!¥cDE! a¬ K6¡´ ± N MVDED !LD L!L K6!LK6!D£EJ a´ " KL ¸ /! K6 D"MNvE ^E ¬ ! EKLN¬ECD¡/KH )QT! KHFE ! EJ!¥c D£E!=·3!%aD!¡QE¡MN

(Y Ntk, ZN

tk) = arg inf

(Y,Z)∈L2(Ftk)E(Y N

tk+1− Y + hf(tk, X

Ntk, Y, Z) − Z∆Wk)

2

L2(Ftk)

N%:JEFE 5E EJ^ ·K6N6KH ! K6 )!D£EJ " E· Ftk

¤ K6EJ$ <ª ¡DE ! £E 3E KE '!"# °d­ ¸ ³ ^´ /MNY$ D

Ytk+1+

∫ tk+1

tk

f(s,Xs, Ys, Zs)ds = Ytk +

∫ tk+1

tk

ZsdWs

# JE d! K6[tk, tk+1]

(Y N

tk)k

E · DK6K6 °7­ ¸ ³ !K6 !D·aD QEJ " JE

Page 43: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³

Page 44: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

t@syj | kpY-~ l t@pY-st@pvmw |w t=|r j.

ª K6K63!MN 3 QEJVD! ! E!!D EJ6K6! MNQE ^´ / ^´

X¥b%!D ·% MNQEJ a´ uD KLE! °c$ ª gD KLE® (XN

tk)0≤k≤N

3 ·-!¡FEHDKLENi

3! Q¡QE"G

XNi,0 = Xi,0

XNi,tk+1

= XNi,tk

+ bi(tk, XNtk

)h+ σi(tk, XNtk

)∆Wk.

^´ /± V) EE:DND)!DD KLE5

X

#/[.$0hKJ , [ .U), [afW ¶S]aU 0fTW \[ , [afW ,CSNhJeg[_W¶UQ]aS¯EC! ¦!: MN MN¡EVD

Φ(X)/ EJ

ΦN(PNtN

)

ΦN aD· /! KL "

(PNtk

)0≤k≤ND EF¡!)I%E -!!K6

d′ ≥ d!NK6DKLEN3! ED !

(XNtk

)0≤k≤N °c! EJ:KL$ 5E: · K6N:!HJEFE )! E5 KLNEED

XNtk

H!- ' K6KEJ C KEJ LFE% D=! -D! K6QEJ 3! !QE!PPEE D!X

OMN P K6KLE J V N¶ · E ¥bDFEJ¬D $QC3MNMN3K6 DD·ECDd = 1

q = 1

G $5` , [^fW UQW\[_]^]aS

±YEC )5aD! ·K6 5!E/JEJ QQEJ6!¦¤EDNLG

Φ(X) = φ(XT ) ¶´ 7 E=QE!LKL5!QEDLD£EJ$\D K6K6N

PNtk

= XNtk

ΦN(PN

tN) = φ(PN

tN) d′

$EJ !1 <± V´ 5E)E MN

φ±YD ¬# 'MNQE # £E /EVD EN¬FED! LK6QE G

E|ΦN(PNtN

) − Φ(X)|2 ≤ Ch

Page 45: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³ $5` , [^fWU-.[^U), [ ¶S

±¶QEC ! !®D· a !E:E D¡!

XGΦ(X) =

φ(XT ,∫ T

0Xtdt)

¶¨ ED PNtk

= (XNtk, h∑k−1

i=0 XNti

)

ΦN(PNtN

) = φ(PNtN

)

d′ ¬ $E !

2 ¥b! aD

φ !QEJ DPD # ! EJKLEJ !:FE'D! ®K6QEJ )E/D

E|ΦN(PNtN

) − Φ(X)|2 ≤ Ch ¶¨ E K6ND ! ¡!E KLEJ ¡ D !6E-K6 6!

X¬QE"K6 5D ! D 3!QEJ ±V

$5` , [^fW ]afvfbU0 Φ(X) = φ(XT ,mint∈[0,T ]Xt,maxt∈[0,T ]Xt)

¨-D a D ¤ ΦN(PN

tN) = φ(PN

tN)

ECDPNtk

= (XNtk,mini≤kX

Nti,maxi≤kX

Nti

) d′

L $E !3 7°c

E# ¬! E KE !FED! LK6QE E|ΦN(PN

tN)−Φ(X)|2 ·¤

K6 KL\ MV¬ D !K6KLNc· ¬KLE ! KLEbE √h log(1/h)

aQE¡K6 QE b¨g) EK6 EJ!5FE-V L!LDND √h

D ! EN VKLEED¬!D KLE)! °c D V adQEJK6 O§ ³ M ³KLEMNMND3 K6$Q# NV´ :3EJ ^E

¨¡E!QE¶D7D EJVD ^´ ! QE\7 MNQE E!(Y N , ZN )! D! Q "EJ E !! D EJ

tk = kh0 ≤ k ≤ N

¨' EJ 3QEJ ¤K6 D MNQE ) E!!D· HEY NtN

= ΦN(PNtN

) J¥c

(Y Ntk, ZN

tk)0≤k≤N−1N3! QE

ZNl,tk

=1

hEk(Y

Ntk+1

∆Wl,k),^´

Y Ntk

= Ek(YNtk+1

) + hf(tk, XNtk, Y N

tk, ZN

tk).

^´ ³KLEMN%MN

Y Ntk

® ! Q!QE ^´ 3D£E@ E DEJ Y 7→ Ek(Y

Ntk+1

) +

hf(tk, XNtk, Y, ZN

tk)

3¡DEJD /!QEJL2(Ftk)

h

)6EK6K6N ¨o 6!D@MN !D QEMN@ EL!@!D EJtkY Ntk

ZNtk

NL! QLE6! EDD! ¨-E KE d!QE FE P -VD ! ! EJ¤ KLEJ ! EDD! 7 EVD Y Ntk

·ZNtk

J­E D· QE ¡ED6 N¡!5V6E " J "LE5D£EJED· EJ !=D(Y N

tk, ZN

tk) °cT E MNQEP# $E 4 !5ª¬ED ¤ ¸ D ¬EJ 5D! ! ^´ T®! !

MN |ZNl,tk

| ≤ 1√h

Ek(Y Ntk+1

)2 c´ ¬ EJ ^EJD 'E5 E K65QE: DD=!

K6N¬MVY Ntk

·ZNtk

EQEN!L2(Ftk)

O E) (Y N

tk, ZN

tk)

!QEgD£EJ!L2 \HE ¡KEMNMNDDL

-! QDK6KL MN'K6K5 ^h

)6EK6K65· !EL2(Ftk)

!'FEMNQE E(Y N

tk+1− Y + hf(tk, X

Ntk, Y, Z) − Z∆Wk)

2.^´

Page 46: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥7´ ³3° @°7\OP¥d°5G<­P´ ¸ ª¬³c´ ¸ Oc´ ¨"¯y°7¯u°I@¥ ¸

¨=PE ¬MNQE DK6K6"N(Y N , ZN )

·(Y, Z)

()ZJNf<h @VegS@A

$ , $ + +- h

+ #$ $ +

max0≤k≤N

E|Ytk − Y Ntk|2 +

N−1∑

k=0

∫ tk+1

tk

E|Zt − ZNtk|2dt

≤ C((1 + |X0|2)h+ E|Φ(X) − ΦN(PN

tN)|2).

:¯J¬77! b¬±¶K6KL !µ¡ EV!DcO c DFE!¬ # Y

!P ENtkE|Ytk −Y N

tk|2 !P: MN EN !P# E

tk+1 E|Ytk+1−Y N

tk+1|2

7 ^´ / a´ QE:!T DYtk − Y N

tk

$ D GYtk − Y N

tk= Etk(Ytk+1

− Y Ntk+1

) + Etk

∫ tk+1

tk

[f(s,Xs, Ys, Zs) − f(tk, XNtk, Y N

tk, ZN

tk)]ds.

°c ·JEE D£EJ v ENP# E 4 )! 7 ^O ¡ QEJN¡# EJD YVN γ > 0

E5D Y E!-G

E|Ytk − Y Ntk|2

≤(1 + γh)E(|Etk(Ytk+1− Y N

tk+1)|2)

+ (h+1

γ)E

∫ tk+1

tk

|f(s,Xs, Ys, Zs) − f(tk, XNtk, Y N

tk, ZN

tk)|2ds

≤(1 + γh)E(|Etk(Ytk+1− Y N

tk+1)|2) + C(h+

1

γ)E

∫ tk+1

tk

|Zs − ZNtk|2ds

+ C(h+1

γ)h(h+ sup

tk≤s≤tk+1

E|Xs −XNtk|2 + sup

tk≤s≤tk+1

E|Ys − Y Ntk|2) ^´

EddFE¡! $E MNf

±Y D ª ! QEJD!P« H2tk,tk+1

(Rq)! VD

φ : [tk, tk+1] → Rq MN 1hE∫ tk+1

tk|φs|2ds <∞ V´ ^EJD !

MNZtk = 1

hEtk

∫ tk+1

tkZsds

" )!(Zs)tk≤s≤tk+1

L2(Ftk) ⊂ H2

tk,tk+1(Rq)

°c@ EJNE: "!¥cNQEEVD !:D·D V N¡G

E

∫ tk+1

tk

|Zs − ZNtk|2ds = E

∫ tk+1

tk

|Zs − Ztk |2ds+ E

∫ tk+1

tk

|Ztk − ZNtk|2ds

= E

∫ tk+1

tk

|Zs − Ztk |2ds+ hE|Ztk − ZNtk|2. ^´

Page 47: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³³KLEMNMN

Ztk =1

hEtk(Ytk+1

∆Wk) +1

hEtk(∆Wk

∫ tk+1

tk

f(s,Xs, Ys, Zs)ds).

°c@DK6QEEJN3E£D5^´ TVGE(|ZN

tk− Ztk |2) ≤

2

h2

q∑

l=1

E|Etk(Ytk+1− Y N

tk+1∆Wl,k)|2

+2

h2E

(

|Etk(∆Wk

∫ tk+1

tk

f(s,Xs, Ys, Zs)ds)|2)

.^´ E

³KLEMN3E¬MV|Etk(Ytk+1

− Y Ntk+1

∆Wl,k)|2 =|Etk(Ytk+1− Y N

tk+1− Etk(Ytk+1

− Y Ntk+1

)∆Wl,k)|2

≤hEtk(|Ytk+1− Y N

tk+1|2) − |Etk(Ytk+1

− Y Ntk+1

)|2@EMNQEJN# E 4 P!ª¬ED N¤ ¸ D ¬E D! a´ E !V PE GE(|ZN

tk− Ztk |2)

≤ChE(|Ytk+1

− Y Ntk+1

|2) − E(|Etk(Ytk+1− Y N

tk+1)|2) + CE

∫ tk+1

tk

f(s,Xs, Ys, Zs)2ds.

^´ ª K6K63 E 77!QE EN$N K6 −E(|Etk(Ytk+1

−Y Ntk+1

)|2) !E ^´ ' KL EJN)!E: 'D !'# °c D· Ea´ ^´ !QEJ^´ Q VN¡G

E|Ytk − Y Ntk|2

≤(1 + γh)E(|Etk(Ytk+1− Y N

tk+1)|2) + C(h+

1

γ)h(h+ sup

[tk,tk+1]

E|Xs −XNtk|2 + sup

[tk,tk+1]

E|Ys − Y Ntk|2)

+ C(h+1

γ)E

∫ tk+1

tk

|Zs − Ztk |2ds+ C(h+1

γ)E(|Ytk+1

− Y Ntk+1

|2) − E(|Etk(Ytk+1− Y N

tk+1)|2)

+ Ch(h+1

γ)E

∫ tk+1

tk

f(s,Xs, Ys, Zs)2ds.

°c%D EN"!6 Nγ = C

< %D4¤ !P KLQ _E£DD EJNC! G

E|Ytk − Y Ntk|2

≤(1 + Ch)E(|Ytk+1− Y N

tk+1|2) + Ch(h+ sup

tk≤s≤tk+1

E|Xs −XNtk|2 + sup

tk≤s≤tk+1

E|Ys − Y Ntk|2)

+ CE

∫ tk+1

tk

|Zs − Ztk |2ds+ ChE

∫ tk+1

tk

f(s,Xs, Ys, Zs)2ds.

Page 48: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥7´ ³3° @°7\OP¥d°5G<­P´ ¸ ª¬³c´ ¸ Oc´ ¨"¯y°7¯u°I@¥ ¸°cQEDK65!

E|Ys − Y Ntk|2 ≤ 2E|Ys − Ytk |2 + 2E|Ytk − Y N

tk|2 ¶·)gEN

hLEK6KLN3· $# @ EE)¡D GE|Ytk − Y N

tk|2

≤(1 + Ch)E(|Ytk+1− Y N

tk+1|2) + Ch(h+ sup

tk≤s≤tk+1

E|Xs −XNtk|2 + sup

tk≤s≤tk+1

E|Ys − Ytk |2)

+ CE

∫ tk+1

tk

|Zs − Ztk |2ds+ ChE

∫ tk+1

tk

f(s,Xs, Ys, Zs)2ds.

°c%EJ MNQEN "KLK6!µ"¬E !D·3O )=N!D5Gmax

0≤k≤NE|Ytk − Y N

tk|2

≤ CE|Φ(X) − ΦN(PNtN

)|2 + Ch+ Cmaxk

suptk≤s≤tk+1

E|Xs −XNtk|2 + sup

tk≤s≤tk+1

E|Ys − Ytk |2

+ CE

N−1∑

k=0

∫ tk+1

tk

|Zs − Ztk |2ds+ ChE

∫ T

0

f(s,Xs, Ys, Zs)2ds.

´ YE ^EJD !" QMN

maxk

suptk≤s≤tk+1

E|Xs −XNtk|2 + sup

tk≤s≤tk+1

E|Ys − Ytk |2 + hE

∫ T

0

f(s,Xs, Ys, Zs)2ds ≤ Ch.

OP =NQEKLN¡G

max0≤k≤N

E|Ytk − Y Ntk|2 ≤ CE|Φ(X) − ΦN(PN

tN)|2 + Ch+ CE

N−1∑

k=0

∫ tk+1

tk

|Zs − Ztk |2ds.

¥b(!5 N Z=DKH*^´ ¬· ^´

E

∫ tk+1

tk

|Zs − ZNtk|2ds

=E

∫ tk+1

tk

|Zs − Ztk |2ds+ hE|Ztk − ZNtk|2

≤E

∫ tk+1

tk

|Zs − Ztk |2ds+ CE|Ytk+1− Y N

tk+1|2 − E|Etk(Ytk+1

− Y Ntk+1

)|2

+ ChE

∫ tk+1

tk

f(s,Xs, Ys, Zs)2ds.

°c@K6KEk

·3=KE ENE∫ T

0f 2(s,Xs, Ys, Zs)ds

Q/ ¡G E

Page 49: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³

N−1∑

k=0

E

∫ tk+1

tk

|Zs − ZNtk|2ds ≤

N−1∑

k=0

E

∫ tk+1

tk

|Zs − Ztk |2ds+ Ch

+ CN−1∑

k=0

E|Ytk+1− Y N

tk+1|2 − E|Etk(Ytk+1

− Y Ntk+1

)|2

≤N−1∑

k=0

E

∫ tk+1

tk

|Zs − Ztk |2ds+ Ch+ CE|Φ(X) − ΦN(PNtN

)|2

+ CN−1∑

k=0

E|Ytk − Y Ntk|2 − E|Etk(Ytk+1

− Y Ntk+1

)|2 ^´ M DEN¡ D EKLN¡! !DH!QE¡E!:KLK6)! KLKHH!:! v°c E !5 a´ =N$

0 ≤ k ≤ N − 1γ > 0

GCE|Ytk − Y N

tk|2 − E(|Etk(Ytk+1

− Y Ntk+1

)|2)

≤CγhE(|Etk(Ytk+1− Y N

tk+1)|2) + C1(h+

1

γ)h(h+ sup

[tk,tk+1]

E|Xs −XNtk|2 + sup

[tk,tk+1]

E|Ys − Y Ntk|2)

+ C1(h+1

γ)E

∫ tk+1

tk

|Zs − ZNtk|2ds.

¨ ! KLE NQEγ = 3C1

D4¤ !$v¡h

E6· $ \V5a"6EDENC1

GCE|Ytk − Y N

tk|2 − E(|Etk(Ytk+1

− Y Ntk+1

)|2)

≤C1hmaxk

E(|Ytk − Y Ntk|2) + C1h

2 +1

2E

∫ tk+1

tk

|Zs − ZNtk|2ds.

°c@ D· E!QE ^´ M QK6 QD£EJ '$ !GN−1∑

k=0

E

∫ tk+1

tk

|Zs − ZNtk|2ds ≤C

N−1∑

k=0

E

∫ tk+1

tk

|Zs − Ztk |2ds+ Ch+ Cmaxk

E(|Ytk − Y Ntk|2).

OP £# ¶Y

·bZ

b!KKLc!7\bDN \! dQE \QEh

·b ¤ ! E KE ! FE D! K6QEJ ! E QE QEJ∑N−1

k=0 E∫ tk+1

tk|Zs − Ztk |2ds

O#\N)D '5E% $FE L2 !

Zb MN'! ¤ 6!QE QE # \­EJ)LDE$\NV:) ¶´ ·)´

Z)D!J! ! aKEJMN¡O / QO#" K6O :K6! E LKLMN

N−1∑

k=0

E

∫ tk+1

tk

|Zt − Ztk |2dt ≤ C(1 + |X0|2)h.

Page 50: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥7´ ³3° @°7\OP¥d°5G<­P´ ¸ ª¬³c´ ¸ Oc´ ¨"¯y°7¯u°I@¥ ¸∑N−1

k=0 E∫ tk+1

tk|Zs − Ztk |2ds

· EdKLE ¬EcK6KH! ED ! $E ¬D ¤º!DD¶DDFEH! K6 EJ ¤

,'SVegUh ¶S%A $ + +-$ , + $ + $ !# , # "+-! $ ,

+-E|Φ(X) − ΦN(PN

tN)|2 $ # - h , # ∑N−1

k=0 E∫ tk+1

tk|Zs − Ztk |2ds ##$ # # #$ ' # 0

, N #$ # , %$ $ +-'$ ! $ # & #$ ! +- %$'& +-, +-$

L2 %$# E|Φ(X)−ΦN (PNtN

)|2 #$ # 0 $ +

! $ # & #$ ! , +# + $ (Y N , ZN ) #

(Y, Z) %, + $ ', + +# $ $ # '$ ! $ # & #$ !+- +-$ , !)+ ! $ + !)+-, ! , +, $ + + ' ! $ #"& #$ ! , +# + $ #$

\ ! E!Q ^ED !3 d Ed3DEEJDKEJ C L!PN MN

Y Ntk

·ZNtkN3! aD· ! KL !

PNtk

G8hf<`cf .£[ ,C[afW A

$ + +- h + #$

$ +

Y Ntk

= yNk (PNtk

), ZNl,tk

= zNl,k(PNtk

) 0 ≤ k ≤ N

1 ≤ l ≤ q,

a´ C (yNk (·))k

(zNl,k(·))k,l

$ $ ! $ ' +,

¨ ¡E # K5TEDH! EC"EVD 5FE'D! ®K6 QE )QE" aD· ! ·¤K6 !¡FE)JE QQE "! D QE¡!I%EJ C °c6 5dK6 MV3 aDyNk (·)

zNl,k(·)±Y D ª D<! D!D£EJED·:±Y D !HELD! @KL QE ·"!HELD QE)!:I%EJ

PNtk

QEPE !5E5D! - E

8hf<`cf .£[ ,C[afW AF $ , $ + +- '

h+ $ +

|yNk0(x) − yNk0(x′)| +

√h|zNk0(x) − zNk0(x

′)| ≤ C|x− x′| a´ '$ #$#$

k0 ≤ N − 1

9M

Page 51: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³ °7/ EN¬ $E P! d aO !"FEHKK6"^E /MV!EEH! K6 ¤EJ -!= KL¡´ Q=3

E|Y N,k0,xtk

− Y N,k0,x′

tk|2

≤ (1 + γh)

1 − Ch(h+ 1γ)E|Ek(Y

N,k0,xtk+1

− Y N,k0,x′

tk+1)|2 +

Ch(h+ 1γ)

1 − Ch(h+ 1γ)E|XN,k0,x

tk−XN,k0,x′

tk|2

+C(h+ 1

γ)

1 − Ch(h+ 1γ)

(E|Y N,k0,x

tk+1− Y N,k0,x′

tk+1|2 − E|Ek(Y

N,k0,xtk+1

− Y N,k0,x′

tk+1)|2).

°c@D EJNγ = C

h

LEK6KLN3· = ^E¡DENCG

E|Y N,k0,xtk

− Y N,k0,x′

tk|2 ≤ (1 + Ch)E|Y N,k0,x

tk+1− Y N,k0,x′

tk+1|2 + ChE|XN,k0,x

tk−XN,k0,x′

tk|2.

E|XN,k0,xtk

− XN,k0,x′

tk|2 ) 6QE

C|x − x′|2 a EN¡ DEMN!& <OP <% EN3 ±YK6KL!Hµ¡¬EY! DPO L·P# NV)´ )% " EJ

yNk0(·) <±Y¡ EJ √

hzNk0(·)! D ¡!5^´

¤­E)DE) E Y V NQEDEJFEMN5! D QEJ¶# NVL´ @ Q QOP =! (!:D EMNQE!"KL3EVD ! aD±Y D

Page 52: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

t@syj | sjkjkl@monqpYr sut@pvmxw q| 7mwk~t@pvmxwk

O9# E=YgE/VgMV !L °7­ ¸ ³! Q6QE^´ "· a´ ¶·VN )(D5!FE% ´ !! KL H aD· yNk (·)

zNl,k(·))

k:

l 7¨JE5EVD DaD!:# E !P!"D· !QEJDD ¬!"! KL

Q 7¥c6ED K6 ! D QEMN@ Etk0 ≤ k ≤ N − 1

7JE D D E KE -!yNk (·) !QE QEJD¡D·Y! )EQE!aD·

p0,k(·)$VD QEJMV

lVPE KEJ 5!

zNl,k(·)!QE QEJD3D<! PQE¡QE!%aD J

pl,k(·)

OkD QEMN E/!! D Etk¬ EV MN

Y Ntk

· E -! "QE' N a ^´ <­ED! K6 · E <E ¬EVD 3! )EKLE¤ ® KLNDEDE! D)

(Y, Z) ·¡¡EJ !DH!· DQDKLK6¡ $¡:KL:! %)!/D QEMN6 E!5!D · EJ Y¥bDFELJE%TD:MN KH5Q! EJ!'¥cD£EJ!

I!%D QEMN' E)!! D E

tk

##

¨%JE! QHE KEJ =!(Y N , ZN )

!)KLE ·E! ¨"®)D:EV¤ KLEJ (Y N,I,I , ZN,I,I)

±¶KL I

ELMN !® ENH!'!D EJtk%TD·

I EJ3!H¥cDE!%P! Q

Y N,I,Itk

D!I

EJ MN ECJEJN!! QY N,I,Itk

®E! Qb!"E KEJ (Y N,I,I

tj )j≥k+1

®TD·QEJN%!D QEMN aI

EJ !¡¥c D£E! ±¶EH EJYN,i,

N−k−1 ︷ ︸︸ ︷

I, · · · , Itk

E ¬\"KLE Ev£EJD= ! ± EJKLEJ '3 E QEJ K6N¬=EJN3MV @E:¡ %aD -!QEp0,N (·) = ΦN(·) !:# E

tN·3=E

αI,I0,N = 1

¥c !:= Etk

0 ≤ k ≤ N − 1G

Page 53: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³ ¨= EJ

α0,Il,k = 0

0 ≤ l ≤ q

¥b0 ≤ i ≤ I − 1

Q= G(αi,I0,k, · · · , αi,Iq,k)

= arg inf(α0,··· ,αq)

E(|αI,I0,k+1.p0,k+1 + hfk(αi−1,I0,k , · · · , αi−1,I

q,k ) − α0.p0,k −q∑

l=1

αl.pl,k∆Wl,k|2).^´ ) /

¯JE KLEJ Y N,I,Itk

EJ ! Q EY N,I,Itk

= αI,I0,k.p0,kZN,I,Il,tk

QEZN,I,Il,tk

=

αI,Il,k .pl,k ¨g)KEJMN# EQEJ 5E£D ^´ bª K6K66 E£! !@! Y) !QE¶D7QE D·7D£EJED· EJ "!

(Y N , ZN )Y! Y7E KLEJ ¯3!QEEHQEJ ´´DK6KLN # EDEED· EJ '!

(Y N , ZN )DK6KLP¦¤ EDD!

± N

(Y N , ZN )·# E KLE D %!H# E¡

(Y N,I,I , ZN,I,I) ¦¤ !QE3 KL" JEN¡G

()ZJNf<h @VegS@AF $ , $ + +- h + #$ $ +

max0≤k≤N

E|Y N,I,Itk

− Y Ntk|2 + h

N−1∑

k=0

q∑

l=1

E|ZN,I,Il,tk

− ZNl,tk

|2

≤ Ch2I−2[1 + |X0|2 + E|ΦN(PN

tN)|2]

+ CN−1∑

k=0

E|Rp0,k(Y N

tk)|2 + Ch

N−1∑

k=0

q∑

l=1

E|Rpl,k(ZN

l,tk)|2.

ª K6K63MN !DLQEpl,k

0 ≤ l ≤ q

DKHNE a!'KL:!" EJ E!$ v±Y:K6ENP¡MN:D¡"aMNHHKLK6PE£DHHDE

CMV HEPMN

N!%" )Q ª ¡KLEJ EJKL D ! K6 5!E § c"! E v\! E! 'D KLN:# HDK6KL L

Y N KLE $EJKLNHZN ­ E5QEJ !QE¡ H K6 v H!QEJ¡:DN VY# " EQE V :K6!

E|Rp0,k(Y N,I,I

tk)|2 ·

E|Rpl,k(ZN,I,I

l,tk)|2 G\E#v# & D6!gD !)QEJ!

aD· 3QEDFEJP MN "3D£EJD =K6N!QE !!¡JD·¤ Rpk

KEJd EKLN!QE EE EJ £E αI,I0,k.p0,k

αI,Il,k .pl,k

¯J3 KL

Page 54: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³3° %°c\O¥7° -GO¥7¥7³¨»´ I@O ´ ¨¡¯y­P° \¨"¯ª ´ ¨"¯ ¸KL·3! EQEJ V!DK63 & D"! aD!¡QEJ ª DQE :KHH! EJ!5¥cD£EJ!

I E !-D QEMN5QE"!HK6v MN c )6EJN)MN

I ¡ $EJ !

2MN6HKL

h2I−2[1 + |X0|2 + E|ΦN(PN

tN)|2]

!· $ £E EE 3E/3DK6KL QEE $

± ! EJ:!5FE'6E 6KHFEJ !-D :! K65´ Y±Y! DLV!=# EJ !@E /!@DNED· !@FED· o(pl,k)

·3!¡FEH D !¡EJKL=D QEMN Etk

E/)!a´ ) / ¨uV! @! E!u J EJ QE%FE G AN(X0) = 1 + |X0|2 +

E|ΦN(PNtN

)|2 Q±¶¡ K6"!¡K6!D£EJ ^´ ) /(αi,I0,k, · · · , αi,Iq,k) =

arg inf(α0,··· ,αq)

E(|αI,I0,k+1.p0,k+1 + hfk(αi−1,I0,k , · · · , αi−1,I

q,k ) − α0.p0,k −q∑

l=1

αl.pl,k∆Wl,k|2)

¡ D θi,Ik = arg inf

θE(Y N,I,I

tk+1+ hfk(α

i−1,Ik ) − θ · vk)2

DEN6-D QE)K66!-JEFEθ∗ = (α∗

0,√hα∗

1, · · · ,√hα∗

q) dª KLK6- ! @ K6"!¡K6 !D£EJ $

θi,Ik Q5G

E(vkv∗k)θ

i,Ik = E(vkY N,I,I

tk+1+ hfk(α

i−1,Ik )).

∆Wk EDN 3 ! !QEJN!

(pl,k)0≤l≤qN ^EJD !cMNE"KEJ D

E(vkv∗k)!@E aK6" VD·¤º!FEQEJ P JE5G

E(vkv∗k) =

E(p0,kp∗0,k) 0 0

0

0

0 0 E(pq,kp∗q,k)

OP =N

E(p0,kp∗0,k)α

i,I0,k = E(p0,kY N,I,I

tk+1+ hfk(α

i−1,Ik )),

E(pl,kp∗l,k)α

i,Il,k = 1

hE(pl,kY

N,I,Itk+1

∆Wl,k),

1 ≤ l ≤ q,

D¡MNY MN E$QV=FEH! Q -!Y N,i,Itk

ZN,i,Il,tk

!/G

ZN,i,Il,tk

=1

hPpl,k

(Y N,I,Itk+1

∆Wl,k

),

^´ ) Y N,i,Itk

= Pp0,k

(Y N,I,Itk+1

+ hf(tk, XNtk, Y N,i−1,I

tk, ZN,i−1,I

tk)).

^´ )

Page 55: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³±¶E5!/ K6¡´ : ¡! ¡= E *",CU`7S/G)) K6 QEJ ¡

E|ZN,i,Il,tk

|2 T³KLEMNMV)V a´ ZN,i,Il,tk

DENi ≥ 1

¶°cYLE D£EJ !5 $EJ H!Lª¬ED N¤ ¸ D ¬E L!|Ek(Y

N,I,Itk+1

∆Wl,k)|2 = |Ek([YN,I,Itk+1

− Ek(YN,I,Itk+1

)]∆Wl,k)|2 ≤ h(Ek[Y

N,I,Itk+1

]2 − [Ek(YN,I,Itk+1

)]2).¥cMN

(pl,k)l Ftk

¤ºKLE ¬- EFE · !DNED· ! EJ!¡JD V G

E|ZN,i,Il,tk

|2 =1

h2E[Ppl,k

(Ek[Y

N,I,Itk+1

∆Wl,k])]2 ≤ 1

h2E(Ek[Y

N,I,Itk+1

∆Wl,k])2

≤ 1

h

(E[Y N,I,I

tk+1]2 − E[Ek(Y

N,I,Itk+1

)]2).

^´ )

ª K6K6® E'EPEL T KLE[Ek(Y

N,I,Itk+1

)]2!QEJ*a´ ) ) DDEMN KE !H ¡QE MN

h!=

0

*",CU`7S F G¶L2

Y N,i,Itk

√hZN,i,I

l,tk

\´ d ^ED 5!K6MN ¡JEE E £EJ

Y N,i,Itk

· √hZN,i,I

l,tk

:!'DE ' N $E b¯3) ^E )K6DN L! KL

L2 aK65i, I, k,N

\O D·5Q\)JχN,Ik : Y ∈

L2(Ftk) 7→ Pp0,k

(Y N,I,Itk+1

+ hf(tk, XNtk, Y, ZN,i−1,I

tk))∈ L2(Ftk).

ª FE KLNE|χN,Ik (Y2) −

χN,Ik (Y1)|2 ≤ (Cfh)2E|Y2 − Y1|2

Cf

E)D E"!¡±Y D !f ¥ E¬D MVN

h)6EK6KLN· $v# E D£E

χN,Ik

DNED· E5E- MN5 N%Y N,∞,Itk

∈ L2(Ftk)^EMN

ZN,i,Il,tk

! !%QEJ3!i ≥ 1

<¨"=E/GY N,∞,Itk

= Pp0,k

(Y N,I,Itk+1

+ hf(tk, XNtk, Y N,∞,I

tk, ZN,I,I

tk)),

^´ ) E|Y N,∞,I

tk− Y N,i,I

tk|2 ≤ (Cfh)

2iE|Y N,∞,Itk

|2 ^´ )

MNY N,0,Itk

= 0 OP E ! d6K63! D ¡

i ≥ 1

E|Y N,i,Itk

|2 ≤(1 +1

h)E|Y N,∞,I

tk− Y N,i,I

tk|2 + (1 + h)E|Y N,∞,I

tk|2

≤(1 + Ch)E|Y N,∞,Itk

|2. ^´ ) E± $E 7D ¤º! E K6\VE d

i = 0D£E

Y N,0,Itk

= 0 ¯ !3 KL

E|Y N,∞,Itk

|2 !)QE ! a´ °c/DKHEN $E ! d'^QEEKL·γEHD E!- E Pp0,k

(Y N,I,Itk+1

) = Pp0,k(Ek[Y

N,I,Itk+1

])EDEJD! Pp0,k

VFE ¡!±Y D "!fG

E|Y N,∞,Itk

|2 ≤(1 + γh)E|Ek[YN,I,Itk+1

]|2

+ Ch(h+1

γ)[Ef 2

k (0, · · · , 0) + E|Y N,∞,Itk

|2 + E|ZN,I,Itk

|2].

^´ )

Page 56: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³3° %°c\O¥7° -GO¥7¥7³¨»´ I@O ´ ¨¡¯y­P° \¨"¯ª ´ ¨"¯ ¸°c EN xK6

E|Y N,∞,Itk

|2 E ^´ ) E KLEEJ Ef 2

k (0, · · · , 0) ≤ C(1 + |X0|2)Q VN

E|Y N,∞,Itk

|2 ≤ (1 + γh)

1 − Ch(h+ 1γ)E|Ek[Y

N,I,Itk+1

]|2 +Ch(h+ 1

γ)

1 − Ch(h+ 1γ)

[1 + |X0|2

]

+C(h+ 1

γ)

1 − Ch(h+ 1γ)

(E|Y N,I,I

tk+1|2 − E|Ek[Y

N,I,Itk+1

]|2) ^´ ) M

V@MNh

LEK6KLN3· $ Q°c@D EJN(!5 Nγ = C

= NE|Y N,∞,I

tk|2 ≤ Ch

[1 + |X0|2

]+ (1 + Ch)E|Y N,I,I

tk+1|2 + ChE|Ek[Y

N,I,Itk+1

]|2

≤ Ch[1 + |X0|2

]+ (1 + 2Ch)E|Y N,I,I

tk+1|2 a´ C

ECDH)D EC °c®DENPDHKLEJ =!QE a´ ) E3E£D

i = ITN

E|Y N,I,Itk

|2 ≤ Ch[1+ |X0|2

]+(1+Ch) E|Y N,I,I

tk+1|2 OP E MVENc K6KL !

µ"¬E T!D·- FE)KE EJ sup0≤k≤N E|Y N,I,I

tk|2 ≤ CAN(X0)

<ª D¶EVD !1^´ ) C ^´ ) E) a´ )VD QQE K65 KE aK6'

E|Y N,i,Itk

|2 ·E|ZN,i,I

l,tk|2 G

supI≥1

supi≥0

sup0≤k≤N

(E|Y N,i,Itk

|2 + hE|ZN,i,Il,tk

|2) ≤CAN(X0).^´ ) /

*",CU`7SP G¡KE EJ -ηN,Ik = E|Y N,I,I

tk− Y N

tk|2 ³KLEMNMN

ηN,IN = 0 <¯3 K6E:FEJ- JE¡

0 ≤ k < NG

ηN,Ik ≤(1 + Ch)ηN,Ik+1 + Ch2I−1AN(X0)

+ CE|Rp0,k(Y N

tk)|2 + Ch

q∑

l=1

E|Rpl,k(ZN

l,tk)|2. a´

¨¡EEJv! DKLNY! !c!7D7FE# KEJ max0≤k≤N E|Y N,I,I

tk−

Y Ntk|2 ! !E\ c K67´ °cQEJNb ¶KK6\EJK6¶MN\N\D!

! a´ E ·3= EN¬# KLEJ ^´ ) / Q=ηN,Ik ≤ Ch2I−1AN(X0) + (1 + h)E|Y N,∞,I

tk− Y N

tk|2

= Ch2I−1AN(X0) + (1 + h)E|Rp0,k(Y N

tk)|2 + (1 + h)E|Y N,∞,I

tk− Pp0,k

(Y Ntk

)|2 a´ C=3EC3 %!:FE:!¡ $EJ E: "!¥ NEEJVD ! Pp0,k

GE|Y N,∞,I

tk− Y N

tk|2 = E|Rp0,k

(Y Ntk

)|2 + E|Y N,∞,Itk

− Pp0,k(Y N

tk)|2. a´

Page 57: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³­P¡<E£D¡KK6¬D MN3MN ^´ ) ¬· ^´ ) =3KL

E|ZN,I,Itk

− ZNtk|2 =

q∑

l=1

E|Rpl,k(ZN

l,tk)|2 +

q∑

l=1

E|ZN,I,Il,tk

− Ppl,k(ZN

l,tk)|2

≤q∑

l=1

E|Rpl,k(ZN

l,tk)|2 +

d

h

(E[Y N,I,I

tk+1− Y N

tk+1]2 − E[Ek(Y

N,I,Itk+1

− Y Ntk+1

)]2),

a´ E|Y N,∞,I

tk− Pp0,k

(Y Ntk

)|2 ≤ (1 + γh)E|Ek[YN,I,Itk+1

− Y Ntk+1

]|2

+ Ch(h+1

γ)[E|Y N,∞,I

tk− Y N

tk|2 + E|ZN,I,I

tk− ZN

tk|2].

a´ °c) D· E a´ v!QE ^´ ) )D E

γ = Cdb) E ^´ ) E

(1 − Ch)E|Y N,∞,Itk

−Pp0,k(Y N

tk)|2 ≤ (1 + Ch)ηN,Ik+1

+ Ch

q∑

l=1

E|Rpl,k(ZN

l,tk)|2 + ChE|Rp0,k

(Y Ntk

)|2. a´ KE¨'DK6 PEJFE!*a´ ) 7- D· E¬D! P $EJ 3!QE%^´ ) C *",CU`7S LgG<I%E EJ-!

ζN = h∑N−1

k=0 E|ZN,I,Itk

− ZNtk|2 ¯K6

ζN ≤Ch2I−2AN(X0) + Ch

N−1∑

k=0

q∑

l=1

E|Rpl,k(ZN

l,tk)|2

+ C

N−1∑

k=0

E|Rp0,k(Y N

tk)|2 + C max

0≤k≤N−1ηN,Ik .

75^´ ) EζN ≤ h

∑N−1k=0

∑ql=1 E|Rpl,k

(ZNl,tk

)|2+d∑N−1k=0

(E[Y N,I,I

tk−Y N

tk]2−E[Ek(Y

N,I,Itk+1

−Y Ntk+1

)]2).

O QE !a´ C ^´ ) =NE|Y N,I,I

tk−Y N

tk|2 − E[Ek(Y

N,I,Itk+1

− Y Ntk+1

)]2 ≤ Ch2I−1AN(X0)

+ CE|Rp0,k(Y N

tk)|2 + [(1 + h)(1 + γh) − 1]E|Ek[Y

N,I,Itk+1

− Y Ntk+1

]|2

+ Ch(h+1

γ)[E|Y N,∞,I

tk− Y N

tk|2 + E|ZN,I,I

tk− ZN

tk|2].

°c D EJN)KE NQEJNγ = 4Cd

h

)6EJKLK6H ))MNdC(h + 1

γ) ≤ 1

2

E£3E¬ ζN ≤ Ch2I−2AN(X0) + Ch

N−1∑

k=0

q∑

l=1

E|Rpl,k(ZN

l,tk)|2 + C

N−1∑

k=0

E|Rp0,k(Y N

tk)|2

+C max0≤k≤N−1

ηN,Ik +1

2h

N−1∑

k=0

E|Y N,∞,Itk

− Y Ntk|2 +

1

2ζN .

Page 58: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³3° %°c\O¥7° -GO¥7¥7³¨»´ I@O ´ ¨¡¯y­P° \¨"¯ª ´ ¨"¯ ¸°c' EE ^´ ) c%a´ ) KEc KL

E|Y N,∞,Itk

−Y Ntk|2 V QE K6N

^´ ) <ª D¶DK6 "E:!= K6¡´ ¤

E

Page 59: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³

Page 60: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

t@syj | r t=qm q| q| mxw1t=|6 syl@vm

ª : EJ:"¡EQEJ !- · E5!'DDc"MN5 KLFEDH!QE¡E H!KLH!-K6!HDE $E

QEH/K6 -KL MNQE /M

KHEJ ! !QE!(PN

tk)0≤k≤N , (∆Wk)0≤k≤N−1

¶ª DcD! !- EJ4K6) N 6!QEJFE6 ¤ºD T­EPD·)VD !T!"aD!D£EρNl,k

ρN,ml,k

$ Q¯3!"!: NT DQ ª ¡D£E VN! KL E/ Y N,i,Itk

, ZN,i,Il,tk

:!*aDDE! !QEJ

Y N,i,I,M,mtk

, ZN,i,I,M,ml,tk

!LE ^E ¡KKL K6 ª PK6 ELN3! !QEJ" EJ EN a @@ E! 8hf<`cf .£[ ,C[afW A

, '$ ! $ +-$ C0

+ #$&- +-$ , + ' $ ! $ (ρNl,k(·) = max(1, C0|pl,k(·)|) : 0 ≤ l ≤ q, 0 ≤ k ≤ N − 1)

#,%,

|Y N,i,Itk

| ≤ ρN0,k(PNtk

),√h|ZN,i,I

l,tk| ≤ ρNl,k(P

Ntk

), p.s.,

' i ≥ 0 I ≥ 0

0 ≤ k ≤ N − 1 O\D J EJ!PE ¤ºD ® VFE! L! aD· _E £EJ c!¤D£EρNl,k

^ ρN,ml,k

¬=! D <°c ¬ MN6G "EN"NEEN

αi,I0,k · p0,k = Y N,i,Itk

l = 0

√hαi,Il,k · pl,k =

√hZN,i,I

l,tk

l ≥ 1^

αi,I0,k · pm0,kl = 0

√hαi,Il,k · pml,k

l ≥ 1

3 3QEJ

2ρNl,k(PNtk

)^

2ρNl,k(PN,mtk

)

! K6" QEJ1

! D!3 aK6 K6N, l, k,m

O N$¶EJPMNQENQ"# "N(Y N,I,I,M , ZN,I,I,M )

·(Y N,I,I , ZN,I,I)

aD· ! KH®!®KHFEJ

M !-QE-! aD -! QE'!®K6

h

Page 61: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³± EJQE d D \DK6 MN 7MN!E § VMN7\ ¶ EJ\! $N¬D EKK63KH P!E DKH ¥b ! KLEJ cEJMN$ aE¡ VMV3D QEMN3QEJ!aD·

pl,k KE FE !

PNtk

E(pl,kp

∗l,k) = Id

§ $CDdNV E TD·dQE\ 7D KE3NK6 MN MN%FE D· ®E%! aD·L - NJEJEN=QEL@EJ aKE £EJ !E:E °c@ Q=! # K6AMk = ∀ j ∈ k, · · · , N − 1 :‖V M

j − Id‖ ≤ h, ‖PM0,j − Id‖ ≤ h ‖PM

l,j − Id‖ ≤ 1

1 ≤ l ≤ q ^´ a) E!'E%¦¤ D 'HE@! g!:KLEJD

V Mj

·PMl,j

¸ %# NV! KLE @D EMNQEJpl,k

3@KE D(V M

k )0≤k≤N−1

(PM

l,k )0≤l≤q,0≤k≤N−1

DNN# !N :E£D6E -6 MVM → ∞ YO#

limM→∞ P(AMk ) = 1

<¯!"!5 NP¡ EJ3 D QEvMNQEJN"!6 & D!=KH"!¡KHEJ $ ()ZJNf<h @VegS@A

$ , $ + +- I ≥ 3 ! + + $ ! $ pl,k $ +-,

E|pl,k|4 <∞ ' k, l h + #$ $ + ' 0 ≤ k ≤ N − 1

E|Y N,I,Itk

− Y N,I,I,Mtk

|2 + h

N−1∑

j=k

E|ZI,Itj − ZI,I,M

tj |2

≤ 9N−1∑

j=k

E(|ρNj (PNtj

)|21[AMk

]c) + ChI−1

N−1∑

j=k

[1 + |X0|2 + E|ρNj (PN

tj)|2]

+C

hM

N−1∑

j=k

(

E‖vjv∗j − Id‖2F E|ρNj (PN

tj)|2 + E(|vj|2|p0,j+1|2)E|ρN0,j(PN

tj)|2

+ h2E[|vj|2(1 + |XN

tj|2 + |p0,j|2E|ρN0,j(PN

tj)|2 +

1

h

q∑

l=1

|pl,j|2E|ρNl,j(PNtj

)|2)])

.

¯"E£PKLN )QEMNH KLK6 MNQE[AM

k ]cDNH

0 MV

M → ∞ KLE ): E£5E: : D ! KLEJ )DDQEN5D'K6-!E# D H!® KL)´ ) T°c ·T# D )! H) KEJ ¡PD)KL! !! NV@ KLNE--'KLK6'! aD·-! QEJ ¥ E-KL ¬ρNj (PN

tj)

E5!3K6KLN! !¡ (!2Q VN¡G

P([AMk ]c) ≤∑N−1

j=k

[P(‖V M

j −Id‖ > h)+P(‖PM0,j−Id‖ > h)+

∑ql=1 P(‖PM

l,j −Id‖ > 1)] C¯EC

P(‖V Mk −Id‖ > h) ≤ h−2E‖V M

k −Id‖2 ≤ h−2E‖V Mk −Id‖2

F = (Mh2)−1E‖vkv∗k−Id‖2F

³KLEMN¡MN5) K65´ -VD555! ¤¦EJVKLMN vOPVD E KL´ ·)´ \ 7KL· a "!LD KEJ K6N¡6QEJ!

Page 62: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @°cbOP¥7° /GI 7«P¨"­° ­° I®¨"¯3° ¤ ª O³3±\¨KL

h<EP! aD3¡KH¡! KHFE EJ !¡ D ! ¯3V!3@@ E!/D¡D -D N!¡QEEKL·$ ´ ¬ 6 $EKLN: '!-KLH EJ6EJVK6MN-!- - K6-!-FE K64DEJ

()ZJNf<h @VegS@AL $ , $ + +- , -

#f !#, +

C1 +- +# (y, z)+ *! ## #$ '$ #$ ,

# #$ $ E|pl,k|2+ε < ∞ k, l ε > 0 , , *! '[√M(αi,I,Mk −

αi,Ik )]i≤I,k≤N−1

! $ #"& +- ,#$ # '$ *! '& + # #$ !#$ , M #$ # , %$ $ 8hS S ¶S]aU-`\h f<`7f .£[ , [afTWA

76E :´ 7NEN7! FE !I%E J :d! E 6K6cMVE" :´ ¡ 3! ^´ ) / D 3E£D3 7 EDD!

Ek

QI@E DE K6¬KLED PE¬D£EJ¬E !QE¬ED· "EPE6YD! Ek

KLEE0

T±Y)®EKLNE ) EN$ ¨" D Y N,i,Itk

= αi,I0,k · p0,k(PNtk

) ­ ®QEJ7QE ^´ ) /:E

CAN(X0) ≥ E|Y N,i,Itk

|2 =

αi,I0,k · E[p0,kp∗0,k]α

i,I0,k ≥ |αi,I0,k|2λmin(E[p0,kp

∗0,k])

­ EQEJ |Y N,i,Itk

| ≤ |αi,I0,k||p0,k(PNtk

)| ≤|p0,k|

CAN(X0)/λmin(E[p0,kp∗0,k])

OP -!ρN0,k(x) = max(1, |p0,k(x)|

CAN(X0)/λmin(E[p0,kp∗0,k])) ­P"KK6

√h|ZN,i,I

l,tk| LE

ρNl,k(x) = max(1, |pl,k(x)|√

CAN(X0)/λmin(E[pl,kp∗l,k])) N³KLEMN

MNHpl,k

)QEH KLE <®Eλmin(E[pl,kp

∗l,k]) = 1

·P D !¡ KL QN$ ¤

8hS S , ZJNf<hK@VegS%A

­PE3E: =

AN,Mk =

1

M

M∑

m=1

|ρN0,k(PN,mtk

)|2, BN,Mk =1

M

M∑

m=1

|fmk (0, · · · , 0)|2.

­P¬KLE · ! 5EE(AN,M

k ) = E|ρN0,k(PNtk

)|2 E(BN,Mk ) ≤ C(1+|X0|2)

JO N$bE ¶FE 7 E!E!!dDNED!E\7DE\! )K67!dK6!D£EJ !QERM EQE EJ/DE

L2(Ω,P) Q¨-D !"" "!"

(xm)1≤m≤M·3

(vm)1≤m≤M!D· !

Rn NEVD !FE"KLEJ DV M = 1

M

∑Mm=1 v

m[vm]∗MN " :

λmin(VM) > 0

TO$# MND·"!Rn θx = arg infθ |x −

θ · v|2M ! ¡QE

θx =[V M ]−1

M

M∑

m=1

vmxm.^´

± EJ D£E x 7→ θx

E P3!¡ $@E:# E 4 λmin(V

M)|θx|2 ≤ |θx.v|2M ≤ |x|2M .^´

Page 63: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³¥bFEH ¡!3D£EJD E5EMN"!

(θi,I,Mk )∗ =((αi,I,M0,k )∗,

√h(αi,I,M1,k )∗, · · · ,

√h(αi,I,Mq,k )∗

)

!5E:FED¡!αi,I,Mk

OP EJ¬!¥c D£E!=!a´ N¡ D

θi+1,I,Mk = arg inf

θ

1

M

M∑

m=1

(ρN,m0,k+1(α

I,I,M0,k+1.p

m0,k+1) + hfmk (αi,I,Mk ) − θ.vmk

)2.

°c NV! EL# K6NAMk

dQE- DK6@FE %!±Y D @!aD·

ρNl,k3@ E# KLE !

pl,kQ=

E|Y N,I,Itk

− Y N,I,I,Mtk

|2 + hN−1∑

j=k

E|ZN,I,Itj − ZN,I,I,M

tj |2 ≤ 9N−1∑

j=k

E(|ρNj (PNtj

)|21[AMk

]c)

+ E(1AMk|αI,I,M0,k − αI,I0,k|2) + h

N−1∑

j=k

q∑

l=1

E(1AMk|αI,I,Ml,j − αI,Il,j |2).

^´ ¥bP E K6H´ ) ®JE'K6 |θI,I,Mk − θI,Ik |2 ·KLN

AMk

Yª D E = E *",CU`7S G\ · !DNED·gFEJ !@FE= (θi,I,Mk )i≥0

b°c H K6 !QE3 "KLK6 JEJN 07SegegS@A

' h + #$ ' AM

k , ' +-$ $ +- + |θi+1,I,M

k − θi,I,Mk |2 ≤ Ch|θi,I,Mk − θi−1,I,Mk |2.

, '$ '$ *! 'θ∞,I,Mk

#,

θ∞,I,Mk = arg inf

θ

1

M

M∑

m=1

(ρN,m0,k+1(α

I,I,M0,k+1.p

m0,k+1) + hfmk (α∞,I,M

k ) − θ.vmk)2.

! $ + |θ∞,I,Mk − θI,I,Mk |2 ≤ [Ch]I |θ∞,I,M

k |2.8hS S

I@E <¥cMN

1 − h ≤ λmin(VMk )

λmax(P

Ml,k ) ≤ 2

0 ≤ l ≤ q

¬AMk

E/V=!5^´ ¬=MN(1 − h)|θi+1,I,M

k − θi,I,Mk |2 ¡QEh2

M

M∑

m=1

(fmk (αi,I,Mk ) − fmk (αi−1,I,M

k ))2

≤Ch2

q∑

l=0

|αi,I,Ml,k − αi−1,I,Ml,k |2λmax(P

Ml,k ) ≤ Ch|θi,I,Mk − θi−1,I,M

k |2.

O E 3 3DE$ ¥bDQ@EMN"E ¬@EEMVθ0,I,Mk = 0

¤

Page 64: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @°cbOP¥7° /GI 7«P¨"­° ­° I®¨"¯3° ¤ ª O³3±\¨*",CU`7S F G3 |θi,I,Mk | 3 ·K6

AMk

<°c@ ·QK6NMV|θi,I,Mk |2 ≤ C

(AN,Mk+1 + hBN,Mk

) AMk

^´ \3! E!QD!

i = ∞ <ª K6KL!QEFEH!=KLK6´ Q=N(1 − h)|θ∞,I,M

k |2 ≤ 1

M

M∑

m=1

[ρN,m0,k+1(α

I,I,M0,k+1.p

m0,k+1) + hfmk (α∞,I,M

k )]2

≤ (1 + γh)AN,Mk+1 + Ch(h+

1

γ)(BN,Mk +

q∑

l=0

|α∞,I,Ml,k |2λmax(P

Ml,k )).

¨®!γ = 8C

h

)6EKLK6N PEJ2C(h + 1

γ)(1 + h) ≤ 1

2(1 − h)

´ $ |θ∞,I,M

k |2 ≤ C(AN,Mk+1 + hBN,Mk ),

KLNEN3MN ^´ JEE i = ∞ T±YKLK6´ ¤ºD D! 3E-EJ!! EJE !

i *",CU`7SG¶E 6!))

θi,I b°cg EN¡FE/ ´ @)EVg!

^´ ) ¤ ´ ) @E:i ≥ 1

|θi,Il,k|2 ≤ E|ρNl,k(PNtk

)|2, 0 ≤ l ≤ q; |θ∞,Ik − θi,Ik |2 ≤ (Cfh)

2iE|ρN0,k(PNtk

)|2. ^´ E³3E ¶EJ# EN7!

θ∞,Ik

£! 7!^´ ) ¤ ´ ) T¶!7 KEJ 4 !¡D QEMNQEpl,k

Gθ∞,Ik = E

(vk[α

I,I0,k+1.p0,k+1 + hfk(α

∞,Ik )]

).

^´ *",CU`7SGLG\­P DK6 !6FE-MNQE

E(1AMk|θI,I,Mk − θI,Ik |2) YO9DEL!KLK6:´

AMk

-N |θ∞,I,Mk − θI,I,Mk |2 ≤ ChI |θ∞,I,M

k |2 ≤ ChI |θ∞,Ik |2 +ChI |θ∞,I,M

k − θ∞,Ik |2.OP = EJN^´ E

E(1AMk|θI,I,Mk − θI,Ik |2) 3 ¡QE

(1 + h)E(1AMk|θ∞,I,Mk − θ∞,I

k |2)

+ 2(1 +1

h)E(1AM

k|θI,I,Mk − θ∞,I,M

k |2) + |θI,Ik − θ∞,Ik |2

≤ (1 + Ch)E(1AMk|θ∞,I,Mk − θ∞,I

k |2) + ChI−1E|ρNk (PNtk

)|2 ^´ M /QE3DK6"!

I ≥ 3 ¸

AMk

V Mk

N 3/3B1 = (Id − (V M

k )−1)θ∞,Ik

B2 = (V Mk )−1

[E(vkρ

N0,k+1(α

I,I0,k+1 · p0,k+1)) −

1

M

M∑

m=1

vmk ρN,m0,k+1(α

I,I0,k+1 · pm0,k+1)

],

B3 = (V Mk )−1h

[E(vkfk(α

∞,Ik )) − 1

M

M∑

m=1

vmk fmk (α∞,I

k )],

B4 =(V M

k )−1

M

M∑

m=1

vmk[ρN,m0,k+1(α

I,I0,k+1 · pm0,k+1) − ρN,m0,k+1(α

I,I,M0,k+1 · pm0,k+1)

+ h(fmk (α∞,Ik ) − fmk (α∞,I,M

k ))].

Page 65: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³OP vQE5a´ ¤º´ P ! DKL

θ∞,Ik − θ∞,I,M

k = B1 + B2 + B3 + B4

YD6MN!AMk

|θ∞,Ik − θ∞,I,M

k |2 ≤ 3(1 +1

h)(|B1|2 + |B2|2 + |B3|2) + (1 + h)|B4|2.

a´ C*",CU`7S OG KE Y E !

B1

B2

B3

B4

AMk

J¨"E 77 EJbDFEMN µ "± M ¡G ¸ ‖Id − F‖ < 1

F−1 − Id =

∑∞k=1[Id − F ]k

‖Id − F−1‖ ≤ ‖F−Id‖1−‖F−Id‖ .

¥ ED MVN

F = V Mk

NE(1AM

k‖Id− (V M

k )−1‖2) ≤ (1−h)−2E‖Id−V Mk ‖2 ≤

(1 − h)−2E‖V Mk − Id‖2

F = (M(1 − h)2)−1E‖vkv∗k − Id‖2F .

O#@EE(|B1|21AM

k) ≤ C

ME‖vkv∗k − Id‖2

F E|ρNk (PNtk

)|2.¥cMN

AMk

‖(V Mk )−1‖ ≤ 2

Q v E(|B2|21AM

k) ≤ C

ME(|vk|2|p0,k+1|2)E|ρN0,k(PN

tk)|2,

E(|B3|21AMk

) ≤ Ch2

ME[|vk|2(1 + |XN

tk|2 + |p0,k|2E|ρN0,k(PN

tk)|2

+1

h

q∑

l=1

|pl,k|2E|ρNl,k(PNtk

)|2)].

ª K6K6!QEFEH!/KLK6P´ ·= EJN ‖PM0,k+1‖ ≤ 1 + h

AMk

/ ^ED K6

(1 − h)|B4|2 ≤ (1 + h)(1 + γh)|αI,I0,k+1 − αI,I,M0,k+1|2 + Ch(h+1

γ)

q∑

l=0

|α∞,Il,k − α∞,I,M

l,k |2.

*",CU`7S G KLEJ QQE ¸ εk = E‖vkv∗k−Id‖2

F E|ρNk (PNtk

)|2+E(|vk|2|p0,k+1|2)E|ρN0,k(PNtk

)|2+h2E[|vk|2(1+|XN

tk|2+

|p0,k|2E|ρN0,k(PNtk

)|2 + 1h

∑ql=1 |pl,k|2E|ρNl,k(PN

tk)|2)].¨ D:" K6 "D ¤º!¡!

B1, B2, B3, B4

!QE ^´ £ v D γ = 3C

hVD ¡!53E3MN

Ch+ Cγ≤ 1

2

QEK6 QDEJ $=NE(1AM

k|θ∞,I,Mk − θ∞,I

k |2) ≤ CεkhM

+ (1 + Ch)E(1AMk|αI,I0,k+1 − αI,I,M0,k+1|2).

I%E %E Vu!u K6K6 ´ ¤ D=·®!%K6 ^´ E ¤ ´ NE(1AM

k|αI,I0,k+1 −

αI,I,M0,k+1|2) ≤ (1 + h)E(1AMk|α∞,I

0,k+1 − α∞,I,M0,k+1 |2) + ChI−1

(1 + |X0|2 + E|ρN0,k+1(P

Ntk+1

)|2 +

E|ρN0,k+2(PNtk+2

)|2).bQEJK6E£K6N

E(1AMk|θ∞,I,Mk − θ∞,I

k |2) ≤ CεkhM

+ ChI−1(1 + |X0|2 + E|ρN0,k+1(P

Ntk+1

)|2

+ E|ρN0,k+2(PNtk+2

)|2)

+ (1 + Ch)E(1AMk|α∞,I,M

0,k+1 − α∞,I0,k+1|2).

Page 66: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @°cbOP¥7° /GI 7«P¨"­° ­° I®¨"¯3° ¤ ª O³3±\¨°c' Ed# EKLNd!DEJD LDK6K63!QE ^´ M N# ! D ∞ ·KLED QE

IdELD QEJ $E _E£D@@D EJN

CJ K6N6!T N ª D ·¡ D

E(1AMk|αI,I,M0,k − αI,I0,k|2) + h

q∑

l=1

E(1AMk|αI,I,Ml,k − αI,Il,k |2)

≤ CεkhM

+ ChI−1(1 + |X0|2 + E|ρN0,k+1(P

Ntk+1

)|2 + E|ρN0,k+2(PNtk+2

)|2)

+ (1 + Ch)E(1AMk|αI,I,M0,k+1 − αI,I0,k+1|2).

¨=DKL "FEH= EN¬" K6KL!)µ"¬EJ ! D3O ) ¤

,'SVegUh ¶S%AF $ +- # , '+-$ !

h $ , "!# +-$

, ! ' , & #$ - I ≥ 3

+ , I ≥ 2

+ +-$ $ $ $ + %, # # +-$ !! $ - $ $ #,%, +-)+# + # , + $ ! $ ' ^´ ) +- , $ +-$ , + !# $ + ,#$ %, , + !' $ ! $ $ !)+ ' ρNl,k +-$

ρN,ml,k

8hS S , ZJVf<hK@VegS:AL±YcEKLN Nc E!E! c V)EN KLN !c!F¤

QD c!J EJ ±Yc! 7K6 !f

QE E !y

zl

NcD K6 ∂0f

·∂lf

±YEEK6·β ∈]0, 1]

3c ! D¬!3« ! ¸ dEJ ! E 3MVε < β

¬MND EMN"QEJ!aD· pl,k

KEJ V¥bk < N − 1

=! Q 3MNQE

AMl,k(α) =1

M

M∑

m=1

vmk ∂lf(tk, XN,mtk

, α0 · pm0,k, · · · , αq · pmq,k)[pml,k]∗,

BMk =

1

M

M∑

m=1

vmk [pm0,k+1]∗, DM

k =√M(Id − V M

k ),

CMk (α) =

M∑

m=1

vmk [αI,I0,k+1 · pm0,k+1 + hfmk (α)] − E

(vk[α

I,I0,k+1 · p0,k+1 + hfk(α)]

)

√M

.

¥bk = N −1

N6BMk = 0

!ECMk (α)

N 7KLαI,I0,k+1 ·pm0,k+1

·αI,I0,k+1 ·p0,k+1N"!5K6FEDD· K6NQE

ΦN(PN,mtN

)

ΦN(PNtN

) <±Y3! !

AMl,k(α)·

DMk

N"PJE !$ v¥b" K6$ D XM w→ E' aN K6%!JEJD /KLEJD

(XM)MDN%^EK5N 3"JEFE "E £E

E HD MNM

!®# )Q# ¥bPFELDND:QE ¡

Page 67: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³LD E b

XM P→ \¥c MNLKHEJ ¡N! !QEb DN ¤D EN3NVE¡G

(AMl,k(αi,Ik ), BM

k , VMk )i≤I−1,l≤q,k≤N−1

P→,

GM = (CMk (αi,Ik ), DM

k )i≤I−1,l≤q,k≤N−1w→ .

^´ /³KLEMN¬MV

limM→∞ V Mk

p.s.= Id

N °c/ £E E aD· f

·ρN,m0,k+1!QE bc!

θi,Ik = E(vk[αI,I0,k+1.p0,k+1+hfk(α

i−1,I0,k , · · · , αi−1,I

q,k )])·θi,I,Mk

! QE ^´ =N|V Mk

√M(θi,I,Mk − θi,Ik ) −DM

k θi,Ik − CM

k (αi−1,Ik )

−BMk

√M(αI,I,M0,k+1 − αI,I0,k+1) − h

q∑

l=0

AMl,k(αi−1,Ik )

√M(αi−1,I,M

l,k − αi−1,Il,k )|

≤ 1k<N−1C√M

|αI,I,M0,k+1 − αI,I0,k+1|2M∑

m=1

|vmk ||pm0,k+1|2

+C√M

|αi−1,I,Mk − αi−1,I

k |1+βM∑

m=1

|vmk ||pmk |1+β.a´

¥b K6¡´ Q=K6"QE DDk

MN([√M(θi,I,Mj − θi,Ij )]j≥k,i≤I ,GM)

w→ <³3EMNθ0,I,Mj = θ0,I

j = 0

j ª ! =! E!

k = N − 1/MN

BMk = 0

·i = 1

OP V! a´ ¤´ CDFE KLN([√M(θi,I,MN−1 − θi,IN−1)]i≤1,GM)

w→ ¥bi = 2

-VMN' KK6PEKLN - E ^´ ¤º´ c·¬'N([√M(θi,I,MN−1 − θi,IN−1)]i≤2,GM)

w→ VMNE¡3 3!QE"^´ DN3 QEJ ¥bdDFEQ= MM = M−1−β/2∑Mm=1 |vmN−1||pmN−1|1+β

= D

1√M

|α1,I,MN−1 − α1,I

N−1|1+βM∑

m=1

|vmN−1||pmN−1|1+β = |√M(α1,I,M

N−1 − α1,IN−1)|1+βMM .

¥cMN[√M(α1,I,M

N−1 −α1,IN−1)]M

d!VJ3ELdE MM D 3 MNM → ∞ c³3KEMN5MN |vN−1||pN−1|1+β ∈ L 2+ε

2+β(P)

OP FE® a -!E!KH$!QEJPD£EJP!EEE EJ# # ! E£D)K6 ) )Q <D! !∑M

m=1 |vmN−1||pmN−1|1+β = O(M2+β2+ε

+r) $ ¬

r > 0 Q¥ EJD MN<=D EN

rLEK6KLN3· $TVN MM → 0

°c@ · EDEKLN$=N([√M(θi,I,MN−1 − θi,IN−1)]i≤I,GM)

w→ ¥b# !D 'MNk < N − 1

@EJ MN"KK6¬D MN$ <´ v/E:¡D ¤=EJ!! P! !BMk

MNY "EJ "DKLK6¡E£JE ¤

Page 68: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @°cbOP¥7° /GI 7«P¨"­° ­° I®¨"¯3° ¤ ª O³3±\¨ # #

±Y@ K6%´ % & ® & D%!%FE ! D EK6'E 6MN® = KL´ MNQEQ=!QE6# 6DK6KL -QE6# E4K5 # & D=!o V%·!KH! aD· !)QE °c®VH! )D D:!(Y N,I,I , ZN,I,I)

(Y, Z)

T DNVN !D! ^E c!N

)Q\! Eb7 V · 7KH! aD·!¡QE!" ¡MN=K63# ! E KLEJ !5# EJNtk

q∑

l=1

E|Rpl,k(√hZN

l,tk)|2 + E|Rp0,k

(Y Ntk

)|2

!¡0

KEJ¬ $EKLN)6EJKLK6V "MNN−1∑

k=0

( q∑

l=1

E|Rpl,k(√hZN

l,tk)|2 + E|Rp0,k

(Y Ntk

)|2) a´ C

!@0 ¬ª DEg@DEL'E£!' )aKE 6 (

yNk (·),√hzNk (·)

)

MNN6 L!QE5E® ´ c°7 QEJ D :HQE¬! DEED·±YD H!:D%aD DV$v E5E"!QEJ"FEQE HNKL MN!QEVD£EJ¡!N! VDH! E δ \§ $¶

aE-!δ

0

!5KLE !'EEJN MN6E-K6KL'^´ £ ! 0

MNN!=3 )Q ³ 7I=N·¤¦ª¬E °c:TcEcMN N

(Y N,I,I,M , ZN,I,I,M )·

(Y N,I,I , ZN,I,I)!:

0E bMV

N!:7 )QV

δ

0V ^E7E)K6N6KHL!'KHEJ

M °cT$bLKLE E! !QE)5 KL´ %DE K63D EN

NKLE $EKLN3DEP!E¡KH! aD·3!E ! D£E ^N®EJ 5!KL

E‖vjv∗j − Id‖2F

E(|vj|2|p0,j+1|2)

°cT$"QE K6H K6!:EQEJ:!H NVDTMN E!EFEQE VK6 MN"!)' EJN ! tk

)d = q = d′ = 1

ª · KLN $ K6EFE aK61[a,b](x)

b− a = δ

Q¨$ DN P!" ¬!QEJ P K6´ H ^EJ !¡ EJ!¬!FEQE

vk KLE EJ :7FE!

PNtk

!FEQE! NVD$ ­EKL E"aD·5KE D!QE !1[a,b](x)

E 1q

P(PNtk

∈[a,b])1[a,b](x)

ª K6K6\KL

E‖vjv∗j−Id‖2F

a b! EDc E! MN2

! DaD· ¡KEJ ¡JEYD c!d# ¶!QE\ KL7´ ! EJ! ! ED3! 1

P(PNtk

∈[a,b])

<¨/KLE PE!KLN " K6MN¡DFEH MNδ → 0

·3MN /! K6!3K6E¬!¡E:QE PMVPNtk

¡¡!QEJ NJEJ !¡δ ­P KLE ® E ¬" -! D= K6! QE/! !D£E D'!T·

E

Page 69: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ QOP±\µ¨"³´¦«I%°¥d¨"©P³±\° ¸ °7­ ¸ ³NVD< E:!LD ! !¡KLE P D ¡ ·JEEJ -!/K6KH!@! /!QEL / K6@´ ·!D=L!6D -!@E£5DK6K6 ^E -E

M aD !KH-! aD:!-QE-5!N

5E5/DND=!# E4K6 ©'!PKE EJ d D "^EJ !DPDE K6Nd ·\ PEQE JEJN

Page 70: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

| n pr | j.sl-t@pv|

i vymxl@pt@kr | r m kp j mql v| ksut@pvmxwkkpT l@|w1t@pv|Yv| 6t=m ~ksx6t=p q| l t=l@myxl@sxq|

Page 71: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 72: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

¯¬EC !QEJ FEQE ´cE · !(Y N

tk, ZN

tk)

! ·PKL K6 !EL2 ! LDE5MVEN " MNQEJ ^´ ¶¯E !/ N " N E@ aK6"! T ED)D! *a´ ¤º´ 3E 63 EJH!'D D= 5!QE:L K6'´ ·: :)!5KLE E ¤ P! "MN KL·N"! ^E JEKH EJ K6QEEK6P!# E4K6 ^QE¬!PKL "·KHP!"aD·¬!"QEJKH! K:FEJ PPE'DD \¨ E' T"DEJ HMV: KLEJ! ¡!QE" K6´ )K6 MNQE N KLKLN ! ! !5! aD !"EJ¤KL PPEL\!

PN MNHD"K6K6 · E NP!6D K6D£E DE 3® E ¯@E -D"@ D=! /!@KLEJ ^EEN/ N/ K6@KH"!%aD!¡QE¡ 3EVD

(Ytk , Ztk)3=KLK6 ¯¬EJ 7!QEJ DPQE 3KLV!QdE! D E LK6MNP¬E£ TD !QEFELQE ´"MNH!·V"VD£EJ D5°7­ ¸ ³x!D·H!HKLE !EMN3!# °7­ ¸ ³ ¥c $<3! D3 E4K5"EVD D°7­ ¸ ³1! DLVDE ¶¯ E EJ¡! DKLK6 !QEJ:# E4K5·5)QH!D:!-# EJ gEJMN'!'D ª E"3# V! @ED@D!¡ !QED

Page 73: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

E

Page 74: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

|-~ l@pvj t@pvmw q| sxvymxl@pt@kr |

#

±¶E'KEJ)!N"¡E£"K6 : EE VH!H# E K6!:EQE)´3 D E "D KEH K6 D43 ! Y N ´ D#VDE)EHQE¬ DE Q¥b¬DE V! 3D

(Y N , ZN)!DKLE :G

Y Ntk

= Etk(YNtk+1

) + hEtkf(tk, XNtk, Y N

tk+1, ZN

tk),

^´´ /hZN

tk= Etk(Y

Ntk+1

∆Wk).^´´

ECDY NtN

= φN(PNtN

) ±¶EL!T DHE£D ^´ ·*a´ PMVH %K6ED !EP !

fY Ntk

QEY Ntk+1

V! EJNE D KLE D d¥bELDN dDZNtk

!QEL/! f

: QEZNtk+1

G 0S ,£SVhegSoS ., SWxS 7S , 2ºfW \UQeSW ,CUQ])`7f ¶hh SVW ¶h S ,Cf ,CS . ]aS .S ., [^egU), [ fTW.W\fW S9NY`b]af .[ S .'`\Uh@h U`\`7f<h , N v°cTD: D6E:!E=D£EJ!" EJYFE)JE ! P!¡FEHKLEMV¡´ ´ <¬E ^ED!K67MVFE"DD6!6 K63´ 3VE E£DDD KE ¥b!E MNEJQEE v)E !@ HVDE ¡ °7­ ¸ ³ ! ¤D ^´´ ·¤ ´´ ¥bHDE)b NV! )LD

RT = (Ri)1≤i≤d′ ¸ :DV6DN"D! !! P!)DEJHELD! =KL QE φN

P!4¤f

a ¤º !6® EJ MN Tª D£ENPKL·! EMNH]aU.f]` , [afW ¶S]¦2 J\U, [^fWohKJ ,Ch f A<h U ¶S®]^f 0U]_[ . JVS=S ., 7f<hWJVS ¨=V! ¬=QE D G

fR(t, x, y, z) = f(t,−R1 ∨ x1 ∧R1, · · · ,−Rd ∨ xd ∧Rd, y, z)

φN,R(x) = φN(−R1 ∨ x1 ∧R1, · · · ,−Rd′ ∨ xd′ ∧Rd′).^´´

¨/! ¡E ‖φN,R‖∞· ‖fR‖∞

MNQE supx |φN,R(x)| ·

supt supx |fR(t, x, 0, 0)|MNV!N!QEJ! 'MN !E 7O QE 6!%D

Page 75: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

-30 -20 -10 0 10 20 30-1

0

1

2

3

4

5

6

7

φN,R φN R1 = 10

!#"$%&(' !")"$ *+*,-" .$/0 !1&2 "$%.43516789)"$ *1&)*:);!"=<#'+#>)<#!"$0 *@?Y N,Rtk

= Etk(YN,Rtk+1

) + hEtkfR(tk, X

Ntk, Y N,R

tk+1, ZN,R

tk), ACB#B $EDF

hZN,Rtk

= Etk(YN,Rtk+1

∆Wk), ACB#B $HGFI+' Y N,R

tN= φN,R(PN

tN)

JLK)M NPORQTSVUXWY[ZTW\^] O;Z_][\`ZXa ZTW\b]cQd Ie <f^*1+<# g'+,h,- ;!!#'ji!+ (Y N,R, ZN,R)

8k +'"l9^<*nmo"$*p 10.=! !+<mC 'rq)"$ < yN,R,Mk (·) / (zN,R,Ml,k (·))1≤l≤q

&s !!#'ji8 Y N,Rtk

/ (ZN,Rl,tk

)1≤l≤q8 yN,R,Mk (PN

tk) /

(zN,R,Ml,k (PNtk

))1≤l≤q

k ,-,-@8 <R.=b8#*"= B /8.$+<;mC '*"= < yN,R,Mk (·) / (zN,R,Ml,k (·))1≤l≤q<* R!1&2 "=<pft8u)"$;!

.=!*<'+&vh'"= )< (αMl,k)0≤l≤q<*R+<Rw89<*+<R!@mC ')"$ < (pl,k)0≤l≤q

x y.zs,-+,-|,h "$>+*h67!y8 <b.zs89#)"$ B /X <#"=,b!.$ M ))~#+'&)"=*<^! (PN ,∆W ) " <#j )<L!|"=<*'*1&)"=<*9)"$ (tk)0≤k≤N

X mo9"$/!**h!.=!<L!*1'+"=</ 9s!.$)L<*!!0q<*@6735 <#"=,b!.$ M )*~#+'*"=#+< (PN

tk,∆Wk,k+1)k

∆Wk,k+1<#` '*+!L.$10)"$*

! %.$+<!*,h"$>+#+<',-<) )<<* %'+ <u)")!1+<%!<%',h<) )+<!;.435" '+#1+,- w*I "= ∆Wk

l<*" PN = XN /"=.<#;<#v-<) !p'+ ! ")*9/fL.3E" <uj tk / ∆Wk<*"=,b!.$+

PNtk+1

! "$<<#"l!&,h.= (PNtk

)∗ =

(

(XNtk

)∗,max0≤i≤kXN,1ti

) y ) XN,1 .z`!*,h"$>+#'+,h<* )`! XN / nI"=*'+*<+/!g<*"$,b!.= PN L,| "=>*@!.$!<gv-'+"$ )/f35**+<p99*"z9w!.=<n.$10)"$*< A '+,h,-L8p&,h!.$/.z-<*"$,b!.z0)" !<*3E nw!*I;q "$ <#!X.3E" )+u9.$.$%!V)+,-!< [tk, tk+1]

Fr~ " <*"4/~) "=',-)V!'+<X9*"=w!.$+<X9.=10)"$*<!"$)"$ ! .$.$+</ <#!!<# 673E <*"=,b!.$RfL'ri!67!n" <uj tk M *1+.="$<)9*"$ <(" !1 ! *+<

D

Page 76: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3°u %­P° ¸ ª¬³3´¦¥c´¨"¯y­P° ± O±\µ¨¡³´ «I@°! oDE £EJ

(∆Wmk,k+1)1≤m≤M

D·MNDN EJoKL K)K D K6 ∆Wm

k

± EJ K6:MN6# EQEJ V5!QED·6QE : D E/KHEJ !6!QE£JE E5!JEJFEJE £EJ MN)!EFE5QE ¡´ D!QEJN! 3D QEJ !D·QE ¤ )-E EJ g!'D·-NV c°7 QE D $¶ KH \SN 2 fT[ .`b]` .¶S UhJ[aUb]^S .UQ]`JNU),£fT[ah S . !D QEMN: EJN

tk O#®:MN !'D QEMN: EN

tk KH

MhKJVU]_[ .U), [afW .

(∆Wmk,k+1)1≤m≤M

> \SgeVegS]afT[ ¶S(∆Wm

k,k+1)1≤m≤M

U S 0o[_W JN`cSVW ¶UQW0S \S . Uh[aU ]aS .(∆Wm

i,i+1,∆Wmi,i+1 : 1 ≤ i ≤ N − 1,1 ≤ m ≤ M)/ ¶¨ JE/ D¡JEJFEJ¡E EJ¡D

Mh JNUQ]^[ .U), [afW . ¶S

PNtk+1

W¶fX, JNS .

(PN,mtk+1

)1≤m≤M

%J EPN,mtk+1

= Tk(PN,mtk

,∆Wmk,k+1)

Tk

! FE:E /!# EJNtk

!5# EJNtk+1

!¡E:D QE ¡!I%EJ CPN

P0

t0 t1 t2 t3 t4 t5

PN,1t1

PN,1t1

PN,2t1

PN,2t1

PN,3t1

PN,3t1

¶ ¸ KHEJ ¬! D KLNEJ FEJ

¨6D c!6 7 aCy(R)

·Cz(R)

MNQd !¡E" VD£E E :! D !QE"D QEJ " D !P3MNY D EE: O QEJ ¬! aD / ψ=! Q 3 aD

[ψ]y·

[ψ]zE¡G

[ψ]y(·) = −Cy(R) ∨ (ψ(·) ∧ Cy(R)),

[ψ]z(·) = −Cz(R) ∨ (ψ(·) ∧ Cz(R)).

± EJ K6!¡DQEJ "¡ N¡ 3%aK6" E!:G→→→ ¨= E

k = NQE

yN,R,MN (·) = φN,R(·) →→→ Ok EJN/! !D EJ

tk

(αMl,k)1≤l≤q =! Q¡DK6KL® K6 K5 aEO H "D -!=K6 K5 ¬=D£EJ!¡KLEJD!¡ /! $ !

E

Page 77: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³KL!¡KL!DE ¡G

infαl

1

M

M∑

m=1

|yN,R,Mk+1 (PN,mtk+1

)∆Wm

l,k

h− αl.p

ml,k|2.

a´´

¨! Q "EzN,R,Ml,k (·) QEJ

zN,R,Ml,k (·) = [αM

l,k.pl,k]z(·) ¥c

αM0,k"! Q DK6KLK6 K6 ¬!=K6"!K6 !DE G

infα0

1

M

M∑

m=1

|yN,R,Mk+1 (PN,mtk+1

) + hf(tk, XN,mtk

, yN,R,Mk+1 (PN,mtk+1

), zN,R,Ml,k (PN,mtk

)) − α0.pm0,k|2.a´´ E

¨=! Q 3Ey

N,R,Mk (·) E

yN,R,Mk (·) = [αM

0,k.p0,k]y(·) →→→ ¨= MN !6# E

t0

¸ T'DK6QEE£DP# E K63! D !QE FEQE´ V'¬ ! D JEJNG O D QEMN EN

tkQ"KL"!K6!DE a´ ^E JEE E £EJ

(PN,mtk

, PN,mtk+1

,∆Wmk )1≤m≤M

E\MN \KLb!K6!\D£E ^´´ ¤´´ E aN N ¬JEE E EJ

(PN,mtk

, PN,mtk+1

,∆Wmk )1≤m≤M

± EJ K6"!D·E ¡ D QE3! EJ !)¥c D£E!@D£EJ3@ ¡!QEJ# MNQE ^´´

Y N,Rtk+1

!QE"! f

­E E4K5!FEQE ´ uJE K K6KL DV6DN(αi,I,Ml,k )0≤l≤q

JEH KLL!'KL!D£EJ '^´ ¥bH!HJ¤ K6¶!dMNQE4¤ D E !dD¶!7FEKLEJDc!7 $ ·b¶!\KL!"DK6 !"FEH '! =KL!¡K6!DE Q !q + 1

KL !K6 ! D£EJ MN ^´´ ¤º´´ EbEQEEJ ^ED MV! ! ¡KLH7MN ^´ ¶°c)Q¶¡MNL ¡KLEJD"!6 !^´´ ¶b!

l!T N N N bKK6\MVE KLNbbKK6\"D ®!5E! aDMN5¡D! ¡!QE¡E'QE HVK6 MN vOP # E K6'!@D@QE /K:=6E E ! K6/FED=NKL MVK6N$ ª !QE¶ EJ4K6)!5E'QE H´3K::!¡!¡ E¡ECD5!"JEJFEJD · ¬QEVK6 MN

­E' L! EJ K67 =!LD£E$ ±¶EEJ-E ®! N!QEP!%D£EJ$ ­PEPE6QE ´ D£EJPPP!)PVD!DN dd# K6[AM

k ]cdMN dKLEJD7! $L 7! K6 3MN

M → ∞ ­PE¬ EJ4K53!PDENFED£EJ ^' FE-VDE EJ P NV EJMVLFE- !6# °7­ ¸ ³! D6! Q 6QE^´´ ¤ ´´ d "·EJMN "DEPDE"! aD ª !E5 MN ¡

E

Page 78: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3°u %­P° ¸ ª¬³3´¦¥c´¨"¯y­P° ± O±\µ¨¡³´ «I@° °c)Q\!gE K5¡D! !@!gEVD )!T N!g K6 ±¶E K6K5 EJ QE ^´ LD D !EJVD /E! K KL%KL! K6 EJ ^´ ) / =K6FE £E# ED¡QE¡K6C¡KL MN

M ¤KHFE ¬!I@·¤¦ª¬E °7=D¡$< v PEVD ! K6! K6EJ VD E MN cPE ­¬¼ LMN3!QEdJD£EN K6 7QE ¤DKLN KLP!'^EJ ¬!=D "K6 !¬D£E $ <± EJ K6P!¡D¡QE P 5E£JE E/!/# !

(Y N , ZN) aKL'! ED£6D! !D MN3DFEJ MN! E KEJ ! EDD! aQEKL µ²¡²¼

±Y3D(Y N

tk, ZN

tk)

! QQEJ"a´´ /c· ^´´ c QEJd V± N ·! E£cVDE # MNQE E!HMN5 :D (Y N,R

tk, ZN,R

tk)

! cQE a´´ P5^´´ P ª" PDMN¡3E ¬D ¤º!$

= I #&- '-&'! # 1:<*#2 # Q-!D MN

(Y N,R, ZN,R)= ¥b-DE JE ! K6-E JEJN5G

8hf<`cf .£[ ,C[afW A A ' h ≤ 1 ∀k : 0 ≤ k ≤ N ∀l : 1 ≤ l ≤ q |Y N,R

tk| ≤ Cy(R)

|ZN,Rl,tk

| ≤ Cy(R)√h

= Cz(R)

Cy(R) = exp(2γ∗ +1 + γ∗

qT )‖φN,R‖2

∞ + 2T (1 +1

γ∗)‖fR‖2

∞ a´´

+ *!γ∗ = 4qC2

f

¯33! E MN3 ±YK6KL"!)µ¡ Ev! D3O 6!D! |Y N,Rtk

| ! |Y N,Rtk+1

| ¥bDE)E MN3 $E ¡! d ^O !L# MNQEJ ^´´ ´ ¶VN$3γ > 0

G

|Y N,Rtk

|2 ≤ (1+γh)|Etk(YN,Rtk+1

)|2+2(h2+h

γ)|fR(tk, X

Ntk, 0, 0)|2+2C2

f (h2+h

γ)Etk|Y N,R

tk+1|+|ZN,R

tk|2,

E E

Page 79: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³ ¡G|Y N,Rtk

|2 ≤(1 + γh)|Etk(YN,Rtk+1

)|2 + 2(h2 +h

γ)|fR(tk, X

Ntk, 0, 0)|2 + 4C2

f (h2 +

h

γ)Etk |Y N,R

tk+1|2

a´´ M + 4C2

f (h2 +

h

γ)

q∑

l=1

|ZN,Rl,tk

|2.

°c- ·JEE'DE P# MNQE ^´´ d·¬-EMNQEJN# $E 4 !"ª¬ED N¤ ¸ D ¬E N $EJ HGh|ZN,R

l,tk|2 ≤ Etk(|Y N,R

tk+1|2) − |Etk(Y

N,Rtk+1

)|2. ^´´ C °c@ D· ED·¡ E !E ^´´ M =N¡G

|Y N,Rtk

|2 ≤(1 + γh)|Etk(YN,Rtk+1

)|2 + 2(h2 +h

γ)|fR(tk, X

Ntk, 0, 0)|2 + 4C2

f (h2 +

h

γ)Etk |Y N,R

tk+1|2

+ 4C2fq(h+

1

γ)Etk(|Y N,R

tk+1|2) − |Etk(Y

N,Rtk+1

)|2. ^´´ /OP =D EN

γ = γ∗ = 4qC2f

!QE ^´´ / Q VN¡G

|Y N,Rtk

|2 ≤ (1 + 2γ∗ +1

qh+

γ∗

qh2)Etk(|Y N,R

tk+1|2) + 2(h2 +

h

γ∗)|fR(tk, X

Ntk, 0, 0)|2.

¥bh ≤ 1

Q= !D5G|Y N,Rtk

|2 ≤ (1 + 2γ∗ +1 + γ∗

qh)Etk(|Y N,R

tk+1|2) + 2h(1 +

1

γ∗)|fR(tk, X

Ntk, 0, 0)|2.

¨="!: PE MN" K6KLP!)µ"¬EJ ! D3O :"G

|Y N,Rtk

|2 ≤ exp(2γ∗+1 + γ∗

qT )Etk(|φN,R(PN

tN)|2)+2h(1+

1

γ∗)Etk

N−1∑

i=k

|fR(ti, XNti, 0, 0)|2.

¨% !L "MN |φN,R(x)| ≤ ‖φN,R‖∞·PMN |fR(ti, x, 0, 0)| ≤ ‖fR‖∞

PECQQEJK6N¡G

|Y N,Rtk

|2 ≤ exp(2γ∗ +1 + γ∗

qT )‖φN,R‖2

∞ + 2T (1 +1

γ∗)‖fR‖2

∞.

¨=! ! 3E ^ED KLN!a´´ C MN |ZN,Rl,tk

| ≤ Cy(R)√h

= Cz(R)

¤

E

Page 80: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3°u %­P° ¸ ª¬³3´¦¥c´¨"¯y­P° ± O±\µ¨¡³´ «I@° = )+& T'-& # V 1:<*#2 #7C !H 3MN # ¬! E¬D·"VDE EJ ¨"'JE: · E EHJ¤ - JE5G8hf<`cf .£[ ,C[afW A A

F ' h + #$

max0≤k≤N

E|Y N,Rtk

− Y Ntk|2 + hE

N−1∑

k=0

|ZNtk− ZN,R

tk|2

≤CE(|φN(PNtN

) − φN,R(PNtN

)|2) + ChE

N−1∑

k=0

|f(tk, XNtk, Y N

tk+1, ZN

tk) − fR(tk, X

Ntk, Y N

tk+1, ZN

tk)|2.

g°7@E MNQE $EJ ! d aO !5E:! D!3 MNQEJ a´´ / a´´ Q VN¡G|Y N,Rtk

− Y Ntk|2 ≤(1 + γh)|Etk(Y

N,Rtk+1

− Y Ntk+1

)|2 + Ch2(1 +1

γh)

(

Etk |f(tk, XNtk, Y N

tk+1, ZN

tk) − fR(tk, X

Ntk, Y N

tk+1, ZN

tk)|2

+ Etk |Y Ntk+1

− Y N,Rtk+1

|2 + |ZNtk− ZN,R

tk|2)

.^´´ C

­PHELKK6 ^E ®MNHP E a´´ C T®PPEL!T D)! MNQEJ^´´ ¬· ^´´ PG

|ZNtk− ZN,R

tk|2 ≤ C

hEtk(|Y N

tk+1− Y N,R

tk+1|2) − |Etk(Y

Ntk+1

− Y N,Rtk+1

)|2. ^´´ C OP @E E

γ V G

|Y N,Rtk

− Y Ntk|2 ≤(1 + Ch)Etk(|Y N

tk+1− Y N,R

tk+1|2)

+ ChEtk |f(tk, XNtk, Y N

tk+1, ZN

tk) − fR(tk, X

Ntk, Y N

tk+1, ZN

tk)|2.

°c%EJ MNQEN ¡±¶K6KL"!)µ"¬E ! D3O )TVNEJ"GE|Y N,R

tk− Y N

tk|2 ≤CE(|φN(PN

tN) − φN,R(PN

tN)|2)

+ ChE

N−1∑

i=k

|f(ti, XNti, Y N

ti+1, ZN

ti) − fR(ti, X

Nti, Y N

ti+1, ZN

ti)|2.

± $E hE∑N−1

k=0 |ZNtk− ZN,R

tk|2 ! ! !3FE¡KK6^E LMNd KL´ <°c=TV a´´ C =N¡G

E M

Page 81: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

hE

N−1∑

k=0

|ZNtk− ZN,R

tk|2

≤CN−1∑

k=0

[E|Y N

tk+1− Y N,R

tk+1|2 − E(|Etk(Y

Ntk+1

− Y N,Rtk+1

)|2)]

≤CE|φN(PNtN

) − φN,R(PNtN

)|2 + CN−1∑

k=0

[E|Y N

tk− Y N,R

tk|2 − E(|Etk(Y

Ntk+1

− Y N,Rtk+1

)|2)]

≤CE|φN(PNtN

) − φN,R(PNtN

)|2 + Cγh+ Ch2(1 +1

γh)

N−1∑

k=0

E|Y Ntk+1

− Y N,Rtk+1

|2

+ Ch2(1 +1

γh)E

N−1∑

k=0

|f(tk, XNtk, Y N

tk+1, ZN

tk) − fR(tk, X

Ntk, Y N

tk+1, ZN

tk)|2

+ Ch2(1 +1

γh)E

N−1∑

k=0

|ZNtk− ZN,R

tk|2

EJN ^´´ C3EL!) E T°c E E! γ = 2C

·"hLEK6N3 E !¡E: -! ! ^EDK6 ¤± N ·/ =MN ! N^EJD K6N'/ KLE EJN'! E D4¤ !' ENL!'KLEEJ 6!%MN-!%! !

XN PN OP o K6 7 6MN /!=6E != MNQEJ E!/VDE E

φNf ³3KLEMN@EJ-MN ! FE KKLKLE MN !QE/FEQE ´

Y N,Rtk

= yN,Rk (PNtk

)·ZN,Rl,tk

= zN,Rl,k (PNtk

)!"aD 3KLE

yN,Rk (·) ·zN,Rl,k (·)

1 ≤ l ≤ q 3± N ·@! / ¤ ! %MN "K6@! D %!

aD· yN,Rk (·) ·

zN,Rl,k (·) MN ¡±YD )aK6 K6NPR

D£EJ aD·φN,R

fRN±¶ D aKL KLN

RG/ ·FE)! K6EJ'!"E) ´ QE 9M $ EJ MN !# !N MN E£D¬! D EJNc!¬±YD

yN,Rk (·) · √hzN,Rl,k (·)! !QE!

R ¨ E PQE !:D5 · H ¡ )¦V5·¡HK:H! aD·!¡QE !: $

Page 82: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

i wksy '| q| syvxmyl%pt@kr |

¯3E ¬!ED¡D QEJ " !5

max0≤k≤N

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2 a´´ /!33! !¡¡#

max0≤k≤N

E

|yN,Rk (x) − yN,R,Mk (x)|2µk(dx)^´´

µk(dx)

! )E6Y!PNtk

T¥7$P! ! !3D¦¤!EN3Z

G

hE

N−1∑

k=0

q∑

l=1

1

M

M∑

m=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2,

hE

N−1∑

k=0

q∑

l=1

|zN,Rl,k (·) − zN,R,Ml,k (·)|2µk(dx).

¯E DKLK6D¬E V!-DE 'KH!EJ MN ¬VN# EQE V"!¡E:QEEJ -!¡# a´´ /

*%(:<+*/:D'- #'++*¨ Fk+1

FEL ! 5QEPJEE E £E (∆Wm

i,i+1)0≤i≤N−1,1≤m≤M·

(∆Wmi,i+1)i≥k+1,1≤m≤M

<¨"/Ek+1 = E(·|Fk+1)

Page 83: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³ :D+\ #*/-'$');+:? #¨=¡QEDY! EJ EJ $

0 ≤ l ≤ q

0 ≤ k ≤ N − 1G

pml,k =pl,k(PN,mtk

),

pml,k+1 =pl,k+1(PN,mtk+1

),

PMl,k =

1

M

M∑

m=1

pml,k(pml,k)

∗.

¨=PE ¬! Q=D EJ/KH¡!¡DV6D 3!¡D !5 Etk

G

PM0,kα

M0,k =

1

M

M∑

m=1

pm0,k

(

[αM0,k+1.pm0,k+1]y + hfR(tk, X

N,mtk

, [αM0,k+1.pm0,k+1]y, [α

Ml,k.p

ml,k]z)

)

,

a´´ PM

0,kα1,M0,k =

1

M

M∑

m=1

pm0,k[αM0,k+1.p

m0,k+1]y,

a´´

PM0,kα

2,M0,k =

h

M

M∑

m=1

pm0,kfR(tk, X

N,mtk

, [αM0,k+1.pm0,k+1]y, [α

Ml,k.p

ml,k]z),

^´´

hPMl,kα

Ml,k =

1

M

M∑

m=1

pml,k∆Wml,k[α

M0,k+1.p

m0,k+1]y,

a´´

PM0,kβ

M0,k =

1

M

M∑

m=1

pm0,kyN,Rk+1 (PN,mtk+1

) + hfR(tk, XN,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

)), ^´´ E

PM0,kβ

1,M0,k =

1

M

M∑

m=1

pm0,kyN,Rk+1 (PN,m

tk+1),

^´´

PM0,kβ

2,M0,k =

h

M

M∑

m=1

pm0,kfR(tk, X

N,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

)),^´´ M

hPMl,kβ

Ml,k =

1

M

M∑

m=1

pml,k∆Wml,ky

N,Rk+1 (PN,m

tk+1).

^´´ C ¸ V b MVEJ bD4¤ ! ! Q N QEbbDVLDN !D ! KLE d MN^ VT KE D

PMl,k

N ! $ D d!P! Q DV6DN¬!D·!¡EHKLE MV ¡!QEJ# E ¡O <¨/" $EKLN-D·!¡DV:¤D !¡JDα

βEk+1(α) = α

Ek+1(β) = β

³KLEMN MV D4¤ ! E! FE EJ fR(t, x, y, zl) = fR(t, x, y, (zl)1≤l≤q)

! KLE ! · /! ·=q

DK6EN!z Q¯MN¡!QEFEH! -!

αMl,N−1

=α0,N .p0,N (·) = φN,R(·)

Page 84: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@° `= $)+*¥baD

ψ=! ¡G

‖ψ‖2k,M =

1

M

M∑

m=1

|ψ(PN,mtk

)|2,

‖ψ‖2k,M

=1

M

M∑

m=1

|ψ(PN,mtk

)|2.

'': #*/- V*¨o! =QE

Kl,k

/D£EJ! QE!pl,k

0 ≤ l ≤ q

d¨o Pl,k

QEJD/D·!aD· \)! QE\FEQEJ

pl,k

[Pl,k]y dK KL7QED ! aD \KLEvMN cE

Cy(R)#

[Pl,k]y = [α.pl,k(·)]y, α ∈ RKl,k ¨H QEKl,k(ω)

0 ≤ k ≤ N−1, 0 ≤

l ≤ q "EJ6!E:KLEJD!¡ =!3 ¬

(pl,k(PN,mtk

)∗)1≤m≤M

+'+7 *¸ :EEK6

βMN

1 < β ≤ 2 ª EEJKL· EFE3V !¬DN D aD5!LEd!K6

h!3 E4K5 ¨LV!Ec E!Ld N EJ !

βKLEd! !¬KEJNQEJNMV

β > 1 DE" MN# E K63D

aV®! MN)E ·MNβ = 2

K6P!)!H !'# E K6L! K KL!'! E!:MN-# H!'!D · EJg K6H! K6¡´ ©a β) <3E=!¡ ·KLN

AMk

! ¶QEJ¡GAMk = ∀ψ ∈ [P0,k+1]y − yN,Rk+1 (·) : ‖ψ‖k+1,M − ‖ψ‖k+1,M ≤ h

β+12 . ^´´ /

`^ )+& )*°c)QvV!K6"! ¡ JEYMN N "!E¡ EQE H!:FEQEE! ! EJ4K6C· b MNb)KLE! DN \ aD!M

h

·3!=KH¡!aD!E5GT1,k,M = ‖yN,Rk (·) − β

M

0,k.p0,k(·)‖2k,M ,

^´´ CT2,k,M = ‖α1,M

0,k − α1,M0,k .p0,k(·)‖2

k,M ,^´´ £

T3,k,M = ‖β2,M0,k − β

2,M

0,k .p0,k(·)‖2k,M ,

^´´ T4,k,l,M = ‖βMl,k.pl,k(·) − zN,Rl,k (·)‖2

k,M ,^´´ C

T5,k,l,M = ‖αMl,k − αMl,k.pl,k(·)‖2k,M .

^´´

Page 85: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³ ¥bdEJQE V7# ^´´ 5EEJ MNc6KLK6¬!Pµ"¬E !D· ´ QdE5!!!FE= JELMV5# !@ EN

tk!@ MN JE !:# E

tk+1

8hf<`cf .£[ ,C[afW A A

E‖yN,Rk (·) − [αM0,k.p0,k]y(·)‖2k,M ≤ET1,k,M + (1 + Ch)E(T2,k,M )

+ (1 + Ch)E‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1]y(·)‖2k+1,M

+C

hET3,k,M + ChE

q∑

l=1

T4,k,l,M + T5,k,l,M.

)bc! E! MN[yN,Rk ]y = yN,Rk

^ )´´ /J:N: ENMN[·]y

1¤ ±¶ D

1

M

M∑

m=1

|yN,Rk (PN,mtk

) − [αM0,k.pm0,k]y|2 ≤

1

M

M∑

m=1

|yN,Rk (PN,mtk

) − αM0,k.pm0,k|2.

¥cQDKLK6*a ^´´

Ek+1

(

yN,Rk+1 (PN,mtk+1

) + hfR(tk, XN,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

))

)

= yN,Rk (PN,mtk

),

βM

0,k

a! Q<QE(^´´ E βM

0,k = Ek+1(βM0,k)

d :!6KL!3K6 !cD£EJ 3G

infα

1

M

M∑

m=1

|yN,Rk (PN,mtk

) − α.pm0,k|2.

­P¡D·¡ · "! KLE / @E MNQE KL"!¥ VQE5G1

M

M∑

m=1

|yN,Rk (PN,mtk

) − αM0,k.pm0,k|2 =

1

M

M∑

m=1

|yN,Rk (PN,mtk

) − βM

0,k.pm0,k|2

+1

M

M∑

m=1

|βM0,k − αM0,k.pm0,k|2

=T1,k,M +1

M

M∑

m=1

|βM0,k − αM0,k.pm0,k|2.

Page 86: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@°° E:! -MN

αM0,k = α1,M0,k + α2,M

0,k

a )H MNQEJa´´ ¤º´´ ¤ ´´ ·βM0,k =

β1,M0,k + β2,M

0,k

5a´´ E ¤ ´´ ¤ ´´ M v ∀γ > 0¶E MVEN E 4 :!

d aO 1

M

M∑

m=1

|βM0,k − αM0,k.pm0,k|2 ≤1 + γh

M

M∑

m=1

|(β1,M

0,k − α1,M0,k ).pm0,k|2

^´´ E

+ (1 +1

γh)

1

M

M∑

m=1

|(β2,M

0,k − α2,M0,k ).pm0,k|2.

\E ¬! E!="K6¬KL"!=K6KH!¡! !5^´´ E ¨MNβ

1,M

0,k .pm0,k

Fk+1

¤ºKLE ¬­ $Ek+1(α

1,M0,k .p

m0,k) = Ek+1(α

1,M0,k ).pm0,k = α1,M

0,k .pm0,k

OP m@E:# $EJ HG

E|β1,M

0,k − α1,M0,k .pm0,k|2 = E|α1,M

0,k − α1,M0,k .pm0,k|2 + E|(α1,M

0,k − β1,M

0,k ).pm0,k|2.°c@K6KEmQ V N¡G

E1

M

M∑

m=1

|β1,M

0,k − α1,M0,k .pm0,k|2 =E

1

M

M∑

m=1

|α1,M0,k − α1,M

0,k .pm0,k|2

+ E1

M

M∑

m=1

|(α1,M0,k − β

1,M

0,k ).pm0,k|2

=E(T2,k,M ) + E1

M

M∑

m=1

|(α1,M0,k − β

1,M

0,k ).pm0,k|2.

ª cY! vK6 cKE EvE¬DNED·"!cFE D· " ((pm0,k)

∗)1≤m≤M!¡FEHKLE JE5G

1

M

M∑

m=1

|(α1,M0,k − β

1,M

0,k ).pm0,k|2 ≤1

M

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2.

¥ E ! N¬EL!K63K6!K6KH!3! ! a´´ KE ¨E)L EN(a+ b)2 ≤ 2(a2 + b2)

G1

M

M∑

m=1

|(β2,M

0,k − α2,M0,k ).pm0,k|2 ≤

2

M

M∑

m=1

|(β2,M

0,k − β2,M0,k ).pm0,k|2

+2

M

M∑

m=1

|(β2,M0,k − α2,M

0,k ).pm0,k|2

=2T3,k,M +2

M

M∑

m=1

|(β2,M0,k − α2,M

0,k ).pm0,k|2.

Page 87: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³¨ LE !@E E@DED· g!E=D·g (

(pm0,k)∗)

1≤m≤ME dMN:DEED6±YD :!

fR^aK6 KLN

R" D a DK6QEE5^´´ ·

^´´ M PG

1

M

M∑

m=1

|(β2,M0,k − α2,M

0,k ).pm0,k|2 ≤Ch2

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

+Ch2

M

q∑

l=1

M∑

m=1

|zN,Rl,k (PN,mtk

) − [αMl,k.pml,k]z|2.

³ D£EJ $ ¯3E£V=FEHKE EJ G

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

≤(1 + γh)1

ME

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2 + ET1,k,M + (1 + γh)E(T2,k,M )

+ 2(1 +1

γh)ET3,k,M + C(h2 +

h

γ)E

1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

+ C(h2 +h

γ)E

1

M

q∑

l=1

M∑

m=1

|zN,Rl,k (PN,mtk

) − [αMl,k.pml,k]z|2.

^´´

¯3E ¬K6 Q D!¡ MNQEJ==3 N EJNE/K6:G

E1

M

q∑

l=1

M∑

m=1

|zN,Rl,k (PN,mtk

) − [αMl,k.pml,k]z|2.

³KLEMN! E!=MN¡D"KL"¡KE "E

E1

M

q∑

l=1

M∑

m=1

|zN,Rl,k (PN,mtk

) − αMl,k.pml,k|2

Page 88: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@°H EN\FE3 ´´ · ^E bMN

[·]z

1¤ ±YD J¥cJ: Eb# E 4

(a+ b+ c)2 ≤ 3(a2 + b2 + c2)Q V N¡G

1

M

M∑

m=1

|zN,Rl,k (PN,mtk

) − αMl,k.pml,k|2 ≤

3

M

M∑

m=1

|zN,Rl,k (PN,mtk

) − βM

l,k.pml,k|2

+3

M

M∑

m=1

|(αMl,k − αMl,k).pml,k|2

+3

M

M∑

m=1

|(βMl,k − αMl,k).pml,k|2

=3T4,k,l,M + 3T5,k,l,M +3

M

M∑

m=1

|(βMl,k − αMl,k).pml,k|2.^´´ M

°c' EJNdEDNED· L!FE¡D· L ((pml,k)

∗)1≤m≤M

· DK6QEJE ^´´ 7·^´´ C Q VN¡G

h

M

M∑

m=1

|(βMl,k − αMl,k).pml,k|2 ≤

1

M

M∑

m=1

|Ek+1(yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y)

∆Wml,k√h

|2

≤ 1

M

M∑

m=1

Ek+1|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

− 1

M

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

^´´ J

@ E $E "!Hª ED N¤ ¸ D ¬E )3E:! $EJ Qª" PD·! KLEEJ -MV¶JE63TK6¡!K6 Q^´´ <°c%T$<% D EJN3D·KE ¤EJ -!QEJ a´´ =!D¡KE

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

E

Page 89: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³QE"G

ET1,k,M + (1 + γh)E(T2,k,M ) + (1 + γh)1

ME

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

^´´ V/+2(1 +

1

γh)ET3,k,M + C(h2 +

h

γ)E

1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

+C(h2 +h

γ)E

q∑

l=1

T4,k,l,M + T5,k,l,M

+C(h+1

γ)E

1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

−C(h+1

γ)E

1

M

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2.

°c®D E !6 γ = C

!QEJP MNQEJ @D ¤º!TYV N$3)D EC! 5G

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

≤(1 + Ch)E1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

+ ET1,k,M + (1 + Ch)ET2,k,M +C

hET3,k,M + ChE

q∑

l=1

T4,k,l,M + T5,k,l,M. ^´´ ¤¥b3E MN¡ K6K6"!)µ"¬E v!D·Q^E!E MN!QE3E5 /´´ )<E

E‖yN,Rk+1 (·)− [αM0,k+1.p0,k+1(·)]y‖2k+1,M

!)EFEJD!E‖yN,Rk+1 (·)− [αM0,k+1.p0,k+1(·)]y‖2

k+1,M

±YPQEE!P !# E"^EJ !# E !3!P# KLN

AMk·¬# Ed EJ d K65G

()ZJNf<h @VegS@ACA

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

≤CE

N−1∑

i=k

T1,i,M + T2,i,M + h−1T3,i,M + Chβ−1

+ CCy(R)2

h

N−1∑

i=k

P([AMi ]c) + ChE

N−1∑

i=k

q∑

l=1

T4,i,l,M + T5,i,l,M.

Page 90: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@°)ª K6KLED bQE$JD K6 cED MND!D· !H K6KL7!µ"¬E ! D E aK6 !QE E ´´ K6E‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2

k+1,M

E‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2

k+1,M

¬¥b-DFE= D ¡GE‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2

k+1,M

=E

(

‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖k+1,M − ‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖k+1,M+

‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖k+1,M

)2

≤(1 + h−1)E

(

‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖k+1,M − ‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖k+1,M

)2

+

+ (1 + h)E‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2k+1,M .

°c ! EN !' ENMV:# ®)HAMk

[AMk ]c

·" ENFEH! Q -!AMk

VN¡GE‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2

k+1,M

≤Chβ +C

hCy(R)2P([AMk ]c) + (1 + h)E‖yN,Rk+1 (·) − [αM0,k+1.p0,k+1(·)]y‖2

k+1,M .

OP ^´´ !V NG

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2 ^´´ J ≤ET1,k,M + (1 + Ch)E(T2,k,M ) + Chβ

+ (1 + Ch)E1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2

+ CCy(R)2

hP([AMk ]c) +

C

hET3,k,M + ChE

q∑

l=1

T4,k,l,M + T5,k,l,M.

©E D£EJ '!=K6K6!µ¡ E K63E!DD ¤

±Y K6´´ ^E 3 N%DE=KH!K63! PMV)PE (! NKLE¬aD· /!

M,h·3!=KH¡!aD !¡QEJ

M

Page 91: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³ = ;)2 #Y-'

T1,k,M07SegegS@A A +# #, $ Kl,k(ω)

# &-$ , +-$'& , + + "! &- # $ $ ,, &-$ $ (pl,k(P

N,mtk

)∗)1≤m≤M

ET1,k,M ≤ CCy(R)2 + ‖fR‖2

∞M

E(K0,k(ω)) + infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2.,'SVegUh ¶S%ACA

+-$ , #$ $ ! , !# $ + $ , +'!#

E(K0,k(ω)) +-K0,k

"+-$ $ $ K0,k(ω) ≤ K0,k #$ +-$ ! $ # #

E(K0,k(ω))+ '$ %$ #

!# !# +-, !)+- +-$ #$ !)+ K0,k(ω) ¿ K0,k $# # $ +- , , + *+ ! $ %$ -"!)+ "! #*! +- δ $ + $ +-$ , + *! $

, + +- $ *!# % $ ! , #$ +-$ , + +- $ ## $ $ + $ ! # δ & #$ # +-$ +-%$# '$ &-)+-$ $ # ! $ ! '$$

K0,k

#, , , #K0,k(ω)

$ %$ - $ %$"+-$ $ $ tk ,,$ # ! ! $ #$+-$ # ', + $ PN,mtk

+! # +-$ , + $ # $ , # ! $ ! $ #$+-$ + # ', + $ $ $ $ # ! , $ $ $',%, +-$ , + + "! "&-# $ $ ! # $ + +-$ , + $ +-$'&

K0,k(ω)

! + #"!

g°c= ENEk+1(β

M0,k) = β

M

0,k

T

ET1,k,M =E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − βM

0,k.pm0,k|2

=E1

M

M∑

m=1

|Ek+1yN,Rk (PN,mtk

) − βM0,k.pm0,k|2

≤E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − βM0,k.pm0,k|2.

¨= PEJ!QE3 "D£E!¡!/ KL¡O ECD

Y = yN,Rk+1 (PNtk+1

) + hfR(tk, XNtk, yNk+1(P

Ntk+1

), zNl,k(PNtk

)),

X = PNtk.

³KLEMNMNE(Y 2|X = x)

KLE QEJ ^@ EJNXNtk

= xPNtk

= x′G

2Cy(R)2 + 2h2E

(

|fR(tk, XNtk, yN,Rk+1 (PN

tk+1), zN,Rl,k (PN

tk))|2∣∣∣∣PNtk

= x′)

≤2Cy(R)2 + 8C2fh

2Cy(R)2 + 8qhC2fCy(R)2 + 4h2|fR(tk, x, 0, 0)|2

≤C(Cy(R)2 + ‖fR‖2∞).

¤

M

Page 92: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@° = ;)2 #Y-'

T2,k,M07SegegS@A AF

ET2,k,M ≤ Cy(R)2

ME(K0,k(ω)).

c=FEH!D ^E !QE# E ¡O /EJ MN @EB

EHKE DP!M

(pm0,k)

∗ bH)HMN B∗BM

= Id\D ! EN)MV

BE@KH"!¡D $EJ !

K0,k(ω) Q­ED¡D£E

α1,M0,k

! QYQE

α1,M0,k =

1

MB∗V

V

! ¡D3!RM !¡DV!

Vm = [αM0,k+1.pm0,k+1]y

<¨@E6E "G

|pm0,k.(α1,M0,k − α1,M

0,k )|2 = (pm0,k)∗B

MV − Ek+1(V )V − Ek+1(V )∗ B

Mpm0,k,

¡G

Ek+1|pm0,k.(α1,M0,k − α1,M

0,k )|2 = (pm0,k)∗B

MEk+1

(

V − Ek+1(V )V − Ek+1(V )∗)B

Mpm0,k.

¨ K6(m,m′)

!¡E:KLEJDEk+1

(V − Ek+1(V )V − Ek+1(V )∗

) JEGEk+1(Vm − Ek+1(Vm)Vm′ − Ek+1(Vm′)) =Ek+1([α

M0,k+1.p

m0,k+1]y[α

M0,k+1.p

m′

0,k+1]y)

− Ek+1([αM0,k+1.p

m0,k+1]y)E

∗k+1([α

M0,k+1.p

m′

0,k+1]y).

¨[αM0,k+1.p

m0,k+1]y = ψ(αM0,k+1, P

N,mtk

,∆Wmk,k+1)

)'D E 5aD gK6Eψ ª K6K6

αM0,k+1

PN,mtk

Fk+1¤ºKLE MNH

m 6= m′ ∆Wmk,k+1

·∆Wm′

k,k+1N3! !QE3! Fk+1

3 ! !QEJN3N¡ <=E/GEk+1([α

M0,k+1.p

m0,k+1]y[α

M0,k+1.p

m′

0,k+1]y) = Ek+1([αM0,k+1.p

m0,k+1]y)Ek+1([α

M0,k+1.p

m′

0,k+1]y)

)!DEk+1(Vm − Ek+1(Vm)Vm′ − Ek+1(Vm′)) = 0

m 6= m′ \OP dK6V¤ !EE)!FEPKLEJD

Ek+1

(V − Ek+1(V )V − Ek+1(V )∗

) N7V$ V&L24S., Yh UQ[FXegSVW , ¶UQW .5] 24UQW\W\]^U, [^fW \S 0S . ,CSNhJeS . <\S@hKJ .£[ ¶S@] 2 [^W , JNh , ¶S%]^U .£[_e \]aU), [ fTW \S .Uh[^Ub]aS.

(∆Wmk,k+1)1≤m≤M

±YKL!EQEJ'KLE NQEEk+1([α

M0,k+1.p

m0,k+1]

2y) ≤ Cy(R)2

M

Page 93: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³OP ‖Ek+1

(V − Ek+1(V )V − Ek+1(V )∗

)‖2 ≤ Cy(R)2 ·3T G

(pm0,k)∗B

MEk+1

(V − Ek+1(V )V − Ek+1(V )∗

) B

Mpm0,k

≤‖Bpm0,kM

‖2‖Ek+1

(V − Ek+1(V )V − Ek+1(V )∗

) B

Mpm0,k‖2

≤‖Bpm0,kM

‖22‖Ek+1

(V − Ek+1(V )V − Ek+1(V )∗

)‖2

≤Cy(R)2

M(pm0,k)

∗B∗B

Mpm0,k

=Cy(R)2

M(pm0,k)

∗pm0,k

=Cy(R)2

M

K0,k(ω)∑

j=1

|pm0,k,j|2.

= E# NV B∗BM

= Id <°c@K6KEJN

mD·¡FEJ VN¡G

Ek+11

M

M∑

m=1

|pm0,k.(α1,M0,k − α1,M

0,k )|2 ≤ Cy(R)2

M

1

M

M∑

m=1

K0,k(ω)∑

j=1

|pm0,k,j|2 =Cy(R)2K0,k(ω)

M

/QE3DK6"! B∗BM

= Id Q´ Y@! D

E(T2,k,M ) ≤ Cy(R)2

ME(K0,k(ω))

¤

= `= ;)2 #Y-'T3,k,M07SegegS@A A

ET3,k,M ≤ C(Cy(R)2 + ‖fR‖∞)h

ME(K0,k(ω)).

g³KLEMVMNβ2,M

0,k − β2,M

0,k = hB∗

MV − Ek+1(V )

ECDFE¡KK63! Q L!B

MN EKLK6 D !V E£D# NV B∗BM

= Id3@ EV

D·!RM !¡DV! G

Vm =fR(tk, XN,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

)).

³KLEMN:E)MNfR(tk, X

N,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

)) = ψ(PN,mtk

,∆Wmk,k+1)

-D EJ aDψ

KLE ±¶5KK65EJK6HMV EK6K6 D ! ¤K6! E 6K66MNEk+1

(Vm − Ek+1(Vm)Vm′ − Ek+1(Vm′)

)= 0

m 6= m′

M

Page 94: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@°OP DK6KL D !K6K6<@E/G

Ek+1

(

(pm0,k)∗B

M

(V − Ek+1(V )V − Ek+1(V )∗

) B

Mpm0,k

)

≤maxm Ek+1(V2m)

M

K0,k(ω)∑

j=1

|pm0,k,j|2.

¨@E6EJ$m

GEk+1(V

2m) ≤Ek+1|fR(tk, X

N,mtk

, yN,Rk+1 (PN,mtk+1

), zN,Rl,k (PN,mtk

))|2

≤2|fR(tk, XN,mtk

, 0, 0)|2 + 4C2fCy(R)2 + 4C2

f

q

hCy(R)2 ≤ C

Cy(R)2 + ‖fR‖2∞

h.

OP v/DDDKLK6¡EJ=KLK6 D !NP "G

ET3,k,M ≤ C(Cy(R)2 + ‖fR‖2

∞)h

ME(K0,k(ω)).

¤

= ;)2 #Y-'T4,k,l,M07SegegS@A A

L

ET4,k,l,M ≤ Cy(R)2

MhE(Kl,k(ω)) +

1

hinfα

E|√hzN,Rl,k (PN

tk) − α.pl,k(P

Ntk

)|2.

¨@E-G

T4,k,l,M =1

M

M∑

m=1

|zN,Rl,k (PN,mtk

) − βM

l,k.pml,k|2 =

1

Mh

M∑

m=1

|√hzN,Rl,k (PN,m

tk) − (

√h)β

M

l,k.pml,k|2.

¨$EJ MNβM

l,k = Ek+1(βMl,k)

3MN √hzN,Rl,k (PN,m

tk)

Fk+1

¤ K6E OP G

E1

Mh

M∑

m=1

|√hzN,Rl,k (PN,m

tk) − (

√h)β

M

l,k.pml,k|2

=E1

Mh

M∑

m=1

|Ek+1√hzN,Rl,k (PN,m

tk) −

√hβMl,k.p

ml,k|2

^´´

≤E1

Mh

M∑

m=1

|√hzN,Rl,k (PN,m

tk) −

√hβMl,k.p

ml,k|2.

M

Page 95: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³¨ )E "!E¡:D£EJ!5! EJ D£E ! K65O ECD

Y =∆Wl,k√

hyN,Rk+1 (PN

tk+1)

·X = PN

tk

<³KLEMN3E ¬MVE(Y 2|X) ≤ Cy(R)2.

OP @E MNQE K6¡O Q V N¡G

ET4,k,l,M ≤ Cy(R)2

MhE(Kl,k(ω)) +

1

hinfα

E|√hzN,Rl,k (PN

tk) − α.pl,k(P

Ntk

)|2.

¤

= ;)2 #Y-'T5,k,l,M07SegegS@A A

O

ET5,k,l,M ≤ Cy(R)2

MhE(Kl,k(ω)).

³EJ MNT5,k,l,M = 1

M

∑Mm=1 |(αMl,k − αMl,k).p

ml,k|2

T­P PD a 3QEB

E:KLEJD"!M

(pml,k)

∗ 3QEJV

D3!RM !¡DV!

Vm =∆Wm

l,k√h

[αM0,k+1.pm0,k+1]y.

¨@E6EJ$ECDD3 EKLEJD G

Ek+1|(αMl,k − αMl,k).pml,k|2 =

1

h(pml,k)

∗B∗

MEk+1

(

V − Ek+1(V )V − Ek+1(V )∗)B

Mpml,k.

³KLEMN E \MN ∆Wml,k√h

[αM0,k+1.pm0,k+1]y = ψ(αM0,k+1, P

N,mtk

,∆Wmk,k+1)

OP# ! FE3K KL^E ¡MN7 D !KLK6N£vK6Y¤ !FEQEP!

Ek+1

(V−Ek+1(V )V−Ek+1(V )∗)N3V$ Q±YK6!EE'KLE NQE

Ek+1(|∆Wm

l,k√h

|2[αM0,k+1.pm0,k+1]

2y) ≤ Cy(R)2.

OP !¡FE:KKL ^E /MV D !K6KLN$<@E/G

ET5,k,l,M ≤ Cy(R)2

MhE(Kl,k(ω)).

¤

M

Page 96: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@° = `^ ;)2 #Y-'

P([AM

k]c)07SegegS@A A

*! , $ "+ $ , +-$ $

P([AMk ]c)

≤2E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

exp

(

− Mhβ+1

144Cy(R)2

)

.

±Y6 EJ)DFEJ MN ¥\ ¡·)))QEKLLE/! ·¤K6EJ ! !QE µ¡²¡²"¼ KL@ \¥bK6 Q¡ J EJ¶! QEJ H FEDE!aD[P0,k+1]y−yN,Rk+1 (·) ´ NV!73 !JEFEE £EJ 5# # !

(Um)1≤m≤MJEEN

1 −1

E£D@QE 12

7! !QE!=E"EE "E EJ KH "!E# E K6 ­P QGZm =PN,m

tk+1

ZM+m =PN,mtk+1

Um = 1

Zm =PN,mtk+1

ZM+m =PN,mtk+1

Um = −1 OP = D:G

P([AMk ]c) =P(∃ψ ∈ H : ‖ψ‖k+1,M − ‖ψ‖k+1,M > h

β+12

)

=P(∃ψ ∈ H :

√√√√ 1

M

M∑

m=1

|ψ(PN,mtk+1

)|2 −

√√√√ 1

M

M∑

m=1

|ψ(PN,mtk+1

)|2 > hβ+1

2

)

=P(∃ψ ∈ H :

√√√√ 1

M

M∑

m=1

|ψ(Zm)|2 −

√√√√ 1

M

M∑

m=1

|ψ(ZM+m)|2 > hβ+1

2

)

D£EJPN,mtk+1

·PN,mtk+1

K KL: Y¨6DJE D6!6FE-! 6 $EJ :D! ¤ QEJNQE(Um)m

´ NV! E¬¡D G = g1, . . . , gn! H E=3 ¬

ψ ∈ H Q v gj ∈ G YMN

√√√√ 1

2M

M∑

m=1

|ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 + |ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 ≤ 1

3√

2h

β+12 .

±YcD£EJ! QE! G \ EE¡KH !cD N2

(

hβ+1

2

3√

2,H, (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M

)

ECD"¬EJ NV! !QE# E"O ³3KEJMN¬MN"DKLK6 P0,k+1

M

Page 97: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³QEDPD·T!P! KL a!D!!K6 ª NDPDE! QE< Q a ± M @ KLO T­P)$ K6! G E

2Cy(R)DFEL · ED£EJ KLN! H ¸

ψ ∈ H 3D! = K6gj ∈ G MNv Q"FED! /!DD ¤º! T¨@E5E$= EN¬# E 4 EFEJ HG

√√√√ 1

M

M∑

m=1

|ψ(Zm)|2 −

√√√√ 1

M

M∑

m=1

|ψ(ZM+m)|2

√√√√ 1

M

M∑

m=1

|ψ(Zm) − gj(Zm)|2 +

√√√√ 1

M

M∑

m=1

|gj(Zm)|2 −

√√√√ 1

M

M∑

m=1

|gj(Zm+M )|2

+

√√√√ 1

M

M∑

m=1

|gj(ZM+m) − ψ(ZM+m)|2.

±Y¡KL K6¡!=K6KH"!¡! KLE QE"G√

2

√√√√ 1

2M

M∑

m=1

|ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 + |ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 ≤ 1

3h

β+12

3E:KK6" $E JEFEJ"! ¬K6¡!=K6KH"!¡! OP G

P(∃ψ ∈ H :

√√√√ 1

M

M∑

m=1

|ψ(Zm)|2 −

√√√√ 1

M

M∑

m=1

|ψ(ZM+m)|2 > hβ+1

2

)

≤P(∃g ∈ G :

√√√√ 1

M

M∑

m=1

|g(Zm)|2 −

√√√√ 1

M

M∑

m=1

|g(ZM+m)|2 > 1

3h

β+12

).

ª ! )QE(PN,m

tk+1, PN,m

tk+1)1≤m≤M

¯3:MN-E%DE G ! ! MNK6N:!DJEFEE EJ¬3E!(Um)m

<¨"/ ¡G

P

(

∃g ∈ G :

√√√√ 1

M

M∑

m=1

|g(Zm)|2 −

√√√√ 1

M

M∑

m=1

|g(ZM+m)|2 > 1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

≤N2

(1

3√

2h

β+12 ,H, (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M

)

maxj

P

(√√√√ 1

M

M∑

m=1

|gj(Zm)|2 −

√√√√ 1

M

M∑

m=1

|gj(ZM+m)|2 > 1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

.

^´´ M

Page 98: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP¯OP± ¸ ° ­P° ± O±\µ¨¡³´ «I@°­P¡KEJ!5EMN# $EJ !«V ! KLEMV3E ¬MVMN

A =

√√√√ 1

M

M∑

m=1

|gj(Zm)|2 −

√√√√ 1

M

M∑

m=1

|gj(ZM+m)|2

@E/GA =

1M

∑Mm=1 |gj(Zm)|2 − |gj(ZM+m)|2

√1M

∑Mm=1 |gj(Zm)|2 +

√1M

∑Mm=1 |gj(ZM+m)|2

≤1M

∑Mm=1 Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2

√1M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

% E3E5! $E ¡MN √a+ b ≤ √

a +√b <O#Q® D· EP!E

^´´ Q/ ¡G

P

(√√√√ 1

M

M∑

m=1

|gj(Zm)|2 −

√√√√ 1

M

M∑

m=1

|gj(ZM+m)|2 > 1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

≤P

( 1M

∑Mm=1 Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2

√1M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

>1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

.

¯JXm = Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2 ¶±\E/DFE G 6! !QE)QE!

(Um)m

'EE(Xm|(PN,m

tk+1, PN,m

tk+1)1≤m≤M ) = 0

°c' |Xm| ≤∣∣|gj(PN,m

tk+1)|2 −|gj(PN,m

tk+1)|2∣∣ VOP =E MN# E 4 P!«VT! a K6¡O C¬·3=N

P

( 1M

∑Mm=1 Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2

√1M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

>1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

≤2 exp

(

−Mhβ+1 1

M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

18 1M

∑Mm=1

(|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2)2

)

.^´´

³ !HEJ aKL "! KL QEJ!¡E aED-!QE3 HG1

M

M∑

m=1

(|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2)2

=1

M

M∑

m=1

(|gj(PN,m

tk+1)| + |gj(PN,m

tk+1)|)2(|gj(PN,m

tk+1)| − |gj(PN,m

tk+1)|)2

≤8Cy(R)2

M

M∑

m=1

(|gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2).

M E

Page 99: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³OP =D EJN!QE ^´´ Q V N

P

( 1M

∑Mm=1 Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2

√1M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

>1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

≤2 exp

(

− Mhβ+1

144Cy(R)2

)

.

¤

M

Page 100: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

s zmxl@sut@pvmxw q| |l=l@|ql ~ mr r pY'| ksywksxvymxl@pt@kr |

O EJ!!cKLE EJ b! KL ! c!:D E D !:N QE K6N K6¡ ENMV¶¡# DK6KL !QE3 EJ4K6()ZJNf<h @VegS@ACA

F

max0≤k≤N

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

≤CCy(R)2 + ‖fR‖2∞

M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + C

N−1∑

k=0

infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2

+ C

N−1∑

k=0

q∑

l=1

infα

E|α.pl,k(PNtk

) −√hzN,Rl,k (PN

tk)|2 + Chβ−1

+C

hCy(R)2e

− Mhβ+1

144Cy(R)2

N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 , (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M))

.

¨- Q"!D¡MNβ

! " (!1

EJEHDND!"# E K6MNh → 0

HMNβ > 2

:N 6D£E) 6KLhβ−1 !V !E :E $£E!·JE

hMN¬ NV NH!QE:E%KE EJ g!-# H!-! D E K6 K6/´ d¨ V!E :E!!QE5D/D E -5# KE ! K:-!D N2

MN¶ NVND ¤º!3KE 3# EJN=E6$ TD!! !¡! K6¡´´ :!33 a´´ 3ZN,R

PNtk¨/JE5 · E K6

Page 101: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³()ZJNf<h @VegS@ACA

max0≤k≤N

E

|yN,Rk (x) − yN,R,Mk (x)|2µk(dx)

≤Chβ−1 + CCy(R)2 + ‖fR‖2

∞M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + C

N−1∑

k=0

infα

E(|yN,Rk (PN

tk) − α.p0,k(P

Ntk

)|2)

+ C

N−1∑

k=0

q∑

l=1

infα

E(|α.pl,k(PN

tk) −

√hzN,Rl,k (PN

tk)|2)

+ CCy(R)2 max0≤k≤N−1K0,k

Mlog(M)

+C

hCy(R)2e

− Mhβ+1

144Cy(R)2

N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 , (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M))

.

O\EN!! K6ND K6<= MN %N3FE5KKL¡MN3 ^´´ /!:/KL¡E!!

CCy(R)2 max0≤k≤N−1K0,k

Mlog(M)

g¯33E!DE:J EJ

‖ψ‖2k =

|ψ(x)|2µk(dx)

"*aDψ

K6E: ¶± E∫|yN,Rk (x) − yN,R,Mk (x)|2µk(dx)

6 ·¤ D EJE‖yN,Rk − yN,R,Mk ‖2

k

=E

(

‖yN,Rk − yN,R,Mk ‖k − 2‖yN,Rk − yN,R,Mk ‖k,M + 2‖yN,Rk − yN,R,Mk ‖k,M)2

≤E

(

max‖yN,Rk − yN,R,Mk ‖k − 2‖yN,Rk − yN,R,Mk ‖k,M , 0 + 2‖yN,Rk − yN,R,Mk ‖k,M)2

≤E(T1) + E(T2)

ECDT1 = 2 max‖yN,Rk −yN,R,Mk ‖k−2‖yN,Rk −yN,R,Mk ‖k,M , 02

T2 = 8‖yN,Rk −yN,R,Mk ‖2k,M

°coEMNQEJN: ' K6-´´ *^EJD !/E(T2)

EN5! /MN D¤QE 5 a´´ / 7¥\5KE E(T1)

coEJ!QE=E=! K6 !E µ²¡²¼ Y= D JEN¡G

P(T1 > u) =P

(

2 max‖yN,Rk − yN,R,Mk ‖k − 2‖yN,Rk − yN,R,Mk ‖k,M , 02 > u

)

≤P

(

∃ψ ∈ [P0,k]y − yN,Rk : ‖ψ‖k − 2‖ψ‖k,M >

√u

2

)

≤3EN2(

√u

24, [P0,k]y − yN,Rk , PN,1:2M

tk) exp(− Mu

576(2Cy(R))2)

a´´ ) /

C

Page 102: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°- E KLO M E)! Q '! N2

^! Q O /dE)J EJPN,1:2Mtk

=

(PN,1tk

, · · · , PN,Mtk

, PN,M+1tk

, · · · , PN,2Mtk

)

(PN,M+mtk

)1≤m≤M 3D ! !QEJN!

(PN,mtk

)1≤m≤M <¨"=E !5 )G

N2(

√u

24, [P0,k]y − yN,Rk , PN,1:2M

tk) =N2(

√u

24, [P0,k]y, P

N,1:2Mtk

)

=N2(

√u

24, [P0,k]y + Cy(R), PN,1:2M

tk).

[P0,k]y +Cy(R)E 7PDFEP!aD· d ·¬ ¬E

2Cy(R) ¨!D6E MV) K6:O aDH KL:"FE' EJ

VG+

"5DE!%aD· G ¬ D ¡MN5G

N2(

√u

24, [P0,k]y + Cy(R), PN,1:2M

tk) ≤3

(2e(2Cy(R))2

u576

log3e(2Cy(R))2

u576

)V([P0,k]y+Cy(R))+

≤3

(1728e(2Cy(R))2

u

)2V([P0,k]y+Cy(R))+

= EMNlog(x) ≤ x

¯3EC3E GV([P0,k]y+Cy(R))+ = V([P0,k]y)+ .°c@ ·Q

n = V([P0,k]y+Cy(R))+·3D ! 3@¦¤ KH !

Rd′+1 EJVD (z1, t1), . . . , (zn, tn).

¥ Eb! Q )!V([P0,k]y+Cy(R))+

b ¤ºKH I

! 1, . . . , n HEMN V g ∈ [P0,k]y + Cy(R)

MNg(zi) ≥ ti

i ∈ I

g(zi) < ti

¨ E®EJ !/D ! ¡:¦¤ KH (z1, t1 − Cy(R)), . . . , (zn, tn − Cy(R)) ¡DD 5MNV([P0,k]y+Cy(R))+ ≤ V([P0,k]y)+

<± E P!QE# E"3$ E !¡FEHK KL"^E ­P¡K KL (^EJD !¡K6N¬MVV([P0,k]y)+ ≤ V(P0,k)+ .

°c9T$:n = V([P0,k]y)+

· D ! 1¦¤ KH !Rd′+1 EVD

(z1, t1), · · · , (zn, tn) MNNb DFE 7QE

([P0,k]y)+ ¥ EJ\! Q £Yb¦¤ KH

I! 1, . . . , n

g ∈ P0,k

QMN[g]y(zi) < ti

i ∈ I

·[g]y(zi) ≥ ti

i /∈ I

´ 6¬!PK6NdMNPi ∈ I g(zi) ≤ [g]y(zi)

i /∈ I g(zi) ≥ [g]y(zi)

= K63!(P0,k)

+ MNY E " ¤ºKH I ¸ MV \

i ∈ IbMN

g(zi) > [g]y(zi) OP DE K6

g(zi) > Cy(R)!D[g]y(zi) = Cy(R)

TOP b!·VE "EC ti > Cy(R)

·" \HE'QEP )! ) K6 ![P0,k]y

MV E E3 dDK6 K6 E 7!i

!QE 1, . . . , n J°7)T$TT K6g′

EJ NDD D E [g′]y(zi) ≥ ti > Cy(R)

DPMN ¬K6 C

Page 103: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³OP

g(zi) ≤ [g]y(zi)

i ∈ I ­ E6QEJvMN

i /∈ IcMN

g(zi) < [g]y(zi) YO¡ DEJKLN

g(zi) < −Cy(R)!D

[g]y(zi) = −Cy(R) QOP#

ti ≤ −Cy(R)·vE6QE3! £EJ\ ¦¤ KH (zi, ti) J°cHT$JDD D EJ ! H KLN

g′!

P0,kMN

[g′]y(zi) < ti ≤ −Cy(R)DMNcK6 O

i /∈ I g(zi) ≥ [g]y(zi) ¨'E!D L K6Nd! P0,k

<MNg(zi) < ti

i ∈ I

g(zi) ≥ ti

i /∈ I

OP V([P0,k]y)+ ≤ V(P0,k)+

ª K6K6""K6 "E:KLEMN"!@E3!E:E5 H!QE µ²¡²¼ Q@EP+

0,k =(z, t) ∈ Rd′ × R : t ≤ g(z) : g ∈ P0,k

⊂(z, t) ∈ Rd′ × R : α.t+ g(z) ≥ 0 : g ∈ P0,k, α ∈ R

.¨@EMN"E KL"O E:DD¡MN

V(P0,k)+ ≤ K0,k + 1.¨==! ! QQE K6# $E 4 )GN2(

√u

24, [P0,k]y + Cy(R), PN,1:2M

tk) ≤ 3

(1728e(2Cy(R))2

u

)2(K0,k+1)

.

°c@ D· E!QE ^´´ ) /T P(T1 > u) ≤9

(1728e(2Cy(R))2

u

)2(K0,k+1)exp

(− Mu

576(2Cy(R))2

)

≤9(12eM)2(K0,k+1) exp(− Mu

2304Cy(R)2

)

u > 576Cy(R)2

M

¨ E73 Vd!3D·!3 E dKE E(T1)

°c@ ·GE(T1) =

∫ ∞

0

P(T1 > u)du

≤v +

∫ ∞

v

P(T1 > u)du

≤v + 9(12eM)2(K0,k+1)

∫ ∞

v

exp(− Mu

2304Cy(R)2

)du

=v + 9(12eM)2(K0,k+1).2304Cy(R)2

Mexp

(− Mv

2304Cy(R)2

)

=φ(v)

v > 576Cy(R)2

M

Q±Y¡KL K)K !E aD φ

PEv =

2304Cy(R)2

Mlog9(12eM)2(K0,k+1).

C

Page 104: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°¥bD¡E$/ ¡G

E(T1) ≤ CCy(R)2K0,k

Mlog(M).

¤

zN,R

Oy MV)E£! ·KL FE:V )!DND)!# yN,R

< ¶^ED 6!LKLFE/V '!D D-!#

zN,R ¨HQEJ D E

()ZJNf<h @VegS@ACAL

hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤Chβ−1 + CCy(R)2 + ‖fR‖2

∞M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + C

N−1∑

k=0

infα

E(|yN,Rk (PN

tk) − α.p0,k(P

Ntk

)|2)

+ C

N−1∑

k=0

q∑

l=1

infα

E(|α.pl,k(PN

tk) −

√hzN,Rl,k (PN

tk)|2)

+C

hCy(R)2e

− Mhβ+1

144Cy(R)2

N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 , (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M))

.

°c@E D /KLEMV¡MN# zN,R

hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

!@KK6"!"MV# yN,R

max0≤k≤N

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2,

D¡MNYE QEY!3K6! °d­ ¸ ³ C

Page 105: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³ a´´ M ¤º´´ !QE N

hE1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤ChE

q∑

l=1

T4,k,l,M + T5,k,l,M

+ CE1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2 − |Ek+1yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2

D¡MN#@KLKLEk

KL·3! D

hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤ChE

N−1∑

k=0

q∑

l=1

T4,k,l,M + T5,k,l,M

+ CN−1∑

k=0

E1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2 − |Ek+1yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2.

°cTD·QEJNoD QEJKLNL! ! D/!QEL /!5KL/!K6KH/!@! cN¡G

hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤ChE

N−1∑

k=0

q∑

l=1

T4,k,l,M + T5,k,l,M

+ C

N−1∑

k=0

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2 − |Ek+1yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2.a´´ )

¨/PE ¬ ^´´ MV¶

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2 − |Ek+1yN,Rk+1 (PN,mtk+1

) − yN,R,Mk+1 (PN,mtk+1

)|2.

C

Page 106: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°´ v^E!D"%a¬D¡V!P ·KLN

(AMk )k °7/T

k ≥ 1G

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

=E‖yN,Rk − yN,R,Mk ‖2k,M

≤(1 + h−1)E

(

‖yN,Rk − yN,R,Mk ‖k,M − ‖yN,Rk − yN,R,Mk ‖k,M)2

+

+ (1 + h)E‖yN,Rk − yN,R,Mk ‖2k,M

≤Chβ +C

hCy(R)2P([AMk−1]

c) + (1 + h)E‖yN,Rk − yN,R,Mk ‖2k,M .

a´´ ) °cDEN/D· ! $EJ !E a´´ ) '= EJN ^´´ QQEJK6N

γ > 0G

hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤ChE

N−1∑

k=0

q∑

l=1

T4,k,l,M + T5,k,l,M + Chβ−1 +C

hCy(R)2

N−1∑

k=0

P([AMk ]c)

+ C max0≤k≤N

E‖yN,Rk − yN,R,Mk ‖2k,M + C(1 + γh)

N−1∑

k=0

E(T2,k,M )

+ Cγh

N−1∑

k=0

1

ME

M∑

m=1

|Ek+1yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2 + C

N−1∑

k=0

ET1,k,M

+ C(1 +1

γh)N−1∑

k=0

ET3,k,M + C(h2 +h

γ)E

N−1∑

k=0

1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

) − [αM0,k+1.pm0,k+1]y|2

+ C(h2 +h

γ)E

N−1∑

k=0

1

M

q∑

l=1

M∑

m=1

|zN,Rl,k (PN,mtk

) − [αMl,k.pml,k]z|2.

³-E !%D γ = 2C

!%h

LEK6KLNH ))EJ aKL6KLE 1M

∑Mm=1 |yN,Rk+1 (PN,m

tk+1) − [αM0,k+1.p

m0,k+1]y|2

MV< NVN7D3!QE7K6KH!! !L# E 4 5D ¤º!b

E 1M

∑Mm=1 |yN,Rk+1 (PN,m

tk+1) − [αM0,k+1.p

m0,k+1]y|2

!=# E!6! · ¤K6AMk

!EKK6(^E -MN%^´´ d QQE K6 QE !)FEDD! K6 ¤

­P¡FEHK KL ^E /MV K6¡´´ ) E ^ED ! E

C

Page 107: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³()ZJNf<h @VegS@ACA

O

hE

N−1∑

k=0

q∑

l=1

|zN,Rl,k (·) − zN,R,Ml,k (·)|2µk(dx)

≤Chβ−1 + CCy(R)2 + ‖fR‖2

∞M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + CN−1∑

k=0

infα

E(|yN,Rk (PN

tk) − α.p0,k(P

Ntk

)|2)

+ CN−1∑

k=0

q∑

l=1

infα

E(|α.pl,k(PN

tk) −

√hzN,Rl,k (PN

tk)|2)

+Cy(R)2Kl,k

Mlog(M)

+C

hCy(R)2e

− Mhβ+1

144Cy(R)2

N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 , (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M))

.

# #

= = *; #2 #.-'+*/'7 +*/-' :D& # #&'

­E¬ KLE E d! ¬MN E)NED E D ! NVN¬ KH!D N2

(h

β+12

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M) b¥b!¬DKL K6 D bD KE EJ N ^E ! c D KH$ ­E" )DE E#®"!HHQE )!HFEL! K6EJ%!® K6H´´ ! !QQEEDEJ -!/ K6¡O MN

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M)

≤3

(216Cy(R)2

hβ+1

)2(K0,k+1+1)

≤C exp(CK0,k+1 logCy(R)

hβ+1).

^´´ )

¨oN6!D=oD /!oK:/!=D@ 3!QEJL /KH/!aD· !¡QE ¨"/! ! 3!5^´´ ) FEHKE EJ -! 3 JEJN5G

C

Page 108: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°

max0≤k≤N

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

+ hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤CCy(R)2 + ‖fR‖2∞

M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + CN−1∑

k=0

infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2

+ CN−1∑

k=0

q∑

l=1

infα

E|α.pl,k(PNtk

) −√hzN,Rl,k (PN

tk)|2 + Chβ−1

+C

h2Cy(R)2 exp

(

CmaxkK0,k log(

Cy(R)

hβ+1) − Mhβ+1

144Cy(R)2

)

.

¯¶T\ vD7 EJ¶!7! KEJ C¯\\ ^E JEJYK KLKLQEEK6h,Kl,k

·M

!KEJ !L3!PD D"EVK6MNFE- !6 °7­ ¸ ³Y !/DKL ) a*a ¡ !Q 5E/! !QED!-E®DKL H!T N:QEEK6 EC6!-K6K5 )h,Kl,k

M

KLEE!¡# $

= = )2 #.*5 & #(+ -'+* ) < #+*¸ 5MN-HED $ !®D QEMN- E:!-! D EJ

tk KLH H! E KEJL!

yN,Rk (·) ·¬!zN,Rl,k (·) aDL!'D L!¬QE

pl,k ª D VE £EJDEcdEC7!7)aKE 7D aD·c DV776QE D $VEKLEMV7!7FE¬ ¤ºD· L ) £¶E£\MN ¶Nb±Y D C¯3¶NC

!PEE NKL MV b! DED\b D · cMN¬EQEpl,k

FE3K KLMN\MV) l ≥ 0

·MNbMN) k O\

0 ≤ k ≤ N − 1

0 ≤ l ≤ q Kl,k = K ¯E3 ! · E

εk(K)FEHK6K6"!3! E KLEJ

infα

E(|yN,Rk (PN

tk) − α.p0,k(P

Ntk

)|2)

+

q∑

l=1

infα

E(|α.pl,k(PN

tk) −

√hzN,Rl,k (PN

tk)|2).

C9E

Page 109: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³±¶E5KLE E -!¡# DK6KL !QE3 EJ4K5¡ D

max0≤k≤N

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2

+ hE

N−1∑

k=0

1

M

M∑

m=1

q∑

l=1

|zN,Rl,k (PN,mtk

) − zN,R,Ml,k (PN,mtk

)|2

≤C(1

N)β−1 + CNK

Cy(R)2 + ‖fR‖2∞

M

+ CN−1∑

k=0

εk(K) +CCy(R)2

h2exp

(

CK log(NCy(R)

T) − MT β+1

144Nβ+1Cy(R)2)

.a´´ )

¨ E \MN MN E4K57D ¬ MNK,N

M

! # )#£^E c! E!5MN

β > 1 MN

maxk εk(K)!¬

0MN

K!:7# QKLE: $EKLN6MN

maxk εk(K)!@)6EK6KLN6V /

06MN ∑N−1

k=0 εk(K)!¡0 ª D¶ QE£EJ@EEE E ¬! ¦!5!QEJ3 K6¡´ ª ¬MVc£ED 7MN )E KLEQE DK6KLN )

M aD H!7EJQEEK65^E@DN- E4K6 d°c · ED! ∑N−1

k=0 εk(K) → 0K6PMNK

D ¬# )DKLK6DEaD K(N)

¥ EJ¬ K6 DE!)E6QEJ)! !DEJD3! D @EL!QEFE5QENK6 MNMN) EEεk(K) ≈ K− 2

d .

´ D7o! 6!D%ECNK− 2

d → 06MV ∑N−1

k=0 εk(K)!=

0 ¨! LE $ E MN3# LEMN EJ K6E£DKH!ED

M<MN

ε(M) =

CK log(NCy(R)

T)− MTβ+1

144Nβ+1Cy(R)2→ −∞ ·¶)6EJKLK6YV vMN

Cy(R)2N2 exp(ε(M)) →0 ­E@DE$dDFE ·V !!

M = CCy(R)2Nβ+1K log(NCy(R)

T)

6DENC

E EJ! ¯d!·E MNK6!3LDKLMNR

! c!d# )^E ¬!3

0# 7!3VDE EJ ¯3d $cD·¬ED· D£E7<FE"VDE EJ )6EK6K6N 'D 3!¬ ! QE$# ! 3¬ £EJ ª D VN!K6 QE K6NMV

M = CNβ+1K log(N)¡D E

CEJ E!

= = `= )+& T< /:< V, #(&UQ] 09b] \S 0fTe` ]aS9N¶[ , J¨oE =ECD=!L!K6

d′, q ·Lo

Kl,k = K5

(l, k) ± EJ K6H

ME D¡!

PN \ª E D· ¡ D NO(MN)

DED ¥c HFE%K6ED'!-# E4K6 \ ! !®D QEJMN/QEH!'K6 aK6) C

Page 110: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°

q+1KLEJDd! $ '

O(MK) EJ $ V¥c $V! ! K6!K6! DE EVD !HDKLEJD ¨-"MV"DK6 N TD·QEJN¬! ! DKL 7JEJ " ¸ "­ d! KLEJD7! $ ±¶EDKL "! ¸ "­y! !®!PE K5KL 3KLE !QE3J¡D£E<@ a ·¤EJN! µ ± M #Q@PDKLDKL

O(MK2)% DQEP# NVMN

M À K OP FEDK6 E!"# !P!

NMK2 §- $¬!DEQE D ¡! aD·!LQEJ\E=DK6 5LK6!vDK6KL5QEJ K6ECD3E"E! VD ! E

δ!QEJ7D QEJMV3NVDN6! T3 Dd!KLK6¬!"MVEN ! !QEN3!3KHEJ ¬MNvNKH !QEDNVD <O# EJN

tk :K66!KL QE¡!6FE-DKL 5 V )E "!6E£¡MN N ED MNcK: !QEc5NVD! DMN7 :K6 D5MN

O(MK2) ¯3¬V!¬ D N '- QE ¯3¬)Q-MNFEDKL !# E K63E£DP KHFE 6! D KLN KLN EJE E

!) ^ED·¬DENKH 3QE E)DK6 !# E K63EJKHEJ L! D ¤K6 KLNE$ ª ¬KHEJ K6 EJ¬D ¬!D QEbD ¬'QE D dED-KLdD MV! ! E D ! !QE !D QEMNQE!"K6$QDK6KL!QE § YQEK6 *h h S\hLSVW 2ºfW0, [^fW ¶S]aU 0fTe` ]aS9N¶[ , J¯¡EJ QEPK6 HDK6K6" P PDK6K6)!E# E ¤ K6 !LFE5DK6 ¯3! D\EP KLK:! ED ! a´´ ) ^DMND!E )EgDE ! ' ¨JE^E 6gE K6)EDK6L!HDEN 7MN'FE=DKL L:D EN

M·HMN'LD

M ≈ Nβ+1+εK

ε > 0EMN® @K6% PE $ E =!·JE/ EK6! aN

K

E 5E! "·)MNL 5KLCNK Cy(R)2+‖fR‖2

M ¡ £EJ !ENhβ−1 E' %!'!

M ≈ Nβ+1+εK·¡EP! KHFE O ^´´ c!MN# dDK6K6!QE# E K6 KLE QE"G

hβ−1 +1

hK2d

,

E LMVFEgDK6 C JE C = Nβ+2+εK3 K = C 1

3N−β+2+ε3

¬°c D· E# E3aD· /!N

C G(

1

N)β−1 +N1+

2(β+2+ε)3d C− 2

3d .

O KHK' !HKL ^ H! DEcN

# E D EN Nbc!:K KL!¡DMNY!6G

N = C 2(3d+2)β+2(2+ε) .

C M

Page 111: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³± 3EVD ¡E

( 1N

)β−1

C− 2(β−1)(3d+2)β+2(2+ε) .

¥ E KLPβ = 2

'N FEKK63V "!PD D¡MNEK6 !5! D ·EJ K6¡·d = 1

a¡ QEHD£E¡ K64ε = 0

DE-!¡3 C− 17

&U-. \S®]^U\U-.S 72 ZY`cSVh 09 cS.6S ,50fTe`bUh UQ[ .fWU S 0:( ?L¯¬E£ ^E dD£EJDD ¤º! !QEJ D£EJ EJT DPQE aD ¤ P!:QEJ Y¯ !QEFELQE )VK6 MN) K6!"E"VD£EJ$ ± K6¡¡ 3K6DY!FE:QE! NVDT$ TO @ E3! <"K6D=!QEJ3E:DK6 !EC$TD QEMNDMNN3 KHFE HMN6KH 6!!QE$ d¥b5D QED=!

M KHFE ! 5E£ EMN7!

KNVD :EQEN Yª DdE- D

Md log(K) ¶±YEJ"K6!FE"DKL Nd £EJ cDE7!QEdD EMNVTDV*^E c6KH! E¤ T ¶E@KH¡! KHFE MN\KH !!QE!DE@ EJ

M EJ3 K:¡!K

D$ O¶!QE3¡D£E!FE5QEJ! NVDFE@DK6 6 E ^ MV KLK6LLKHL! E)H KHL!H!QEJ!L!D EJ-)! !L!NMd log(K)

¨ MV# EL£EJD EEJ3E/D£E EY= K6¡!KL QE!¡FEHDK6 JEJFE NMdK2

°cg EN¡ :KL5log(K)

!QEFE-DK6 v5KK65D£E DcMNLD7TD· !QE3"DE( EY! aβ = 2

¬!¡# !"! C− 14+ε+d

Qε > 0

3MNM ≈ N1+β+εK

Q¯EJ¬DKLE3E£D¡# E K6! D !QE § ¥bD·PEJ K6)!5D QEMN EN!¡!D · EJtk

TD QED!M

E D ¤cKH c ENtk

^E cMNN

QEJMV7!M

EDcN7 K: '! TD·-DEDT KL MVENM

E D· K: T # Etk+1

ª K63 D 3! N $E! ¸ V!@MN¶!6D (!5D£E D !E3¡D£EJ! ! E KLEMNbP aK6 D !EP DEP!%KLV!§ FED¤ ¸ D ¥cFE !E3DDE$<MN¶¡^ECE "DKL ·V!^E:DK6 "!E4K6MNQEK63! !QE!@K6V! " Q±¶E5DKL " PE!L !5!M2N

¨¶# )EJVD !@ E4K65)! L!QE 5 K6 !QE § Q7d!# !! N

2+d4√

M

¸ Q5K6 c*aD :!E"DK6 C− 1

13+d ±¶E®! D-5!D': KL EJNE£D-# E4K6L! D :!QEJD·¡QE ¸ v'3!K6N E ^ED·

2/! EJK6¬PK6!D£EJDY!

213+d !QE"DE § ¶DN24+d+ε D ³KLEMN)QVMNPD EMNPNVDTP! Q !QE VK6FEJ cVD£EJ$FE¬DKL 7D E7QE\E YMNd Y! EJKLEJ " EJKL Cª DVEK6

C

Page 112: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´¦³° ) %I@O"¢¨"³Oc´ ¨"¯­°± °7³3³°7©³ ª¨¡I®I@´ ¸ ° ­OP¯ ¸ ± O±\µ¨¡³´ «I@°D)¡EJDK6 ¡!)E K5 Q¯3 QD K6!QE3E:QEJ "NKL MV

Page 113: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

Page 114: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

n t=|wk-pvmxwk

¯E£ N D gEJ K6HMV7 *!/D QEMNLE!:K6M

D K6N E E (∆Wm

k,k+1)1≤m≤M <±¶E6E=! 3ENK63¤!EEH!FEKLEJD

A = Ek+1

(V − Ek+1(V )V − Ek+1(V )∗

) V

¬Dd!RM !PDV!

([αM0,k+1.pm0,k+1]y

∆Wml,k√h

)1≤m≤M

([αM0,k+1.pm0,k+1]y)1≤m≤M

± EVFEJ!¡K6"¤º!FEQE ®!A

¡K6!5D ‖A‖2

·E !5DN " K6Ek+1

1M

∑Mm=1 |αMl,k − Ek+1(α

Ml,k).pml,k|2

Ek+1

1M

∑Mm=1 |αM0,k − Ek+1(α

M0,k).pm0,k|2a¬EHKE EJ '!K6

T2,k,M

T5,k,l,M

Q°c=T$Q`cf \h0fTW ,Ch]^SNh6]^U`bhKJ .SVW 0S¶S

Z¶UQW ./]^S \h[TSNh ¡E£! E MNPH $E !5ª¬ED N¤ ¸ D ¬E : EJ '^´´ J )65E MN:D= E b6EC6 !-E£JE ECD

Ek+1(αMl,k)

¬αMl,k

³KLEMV MN" EVFEJ L!¬KL¬¤ !EQE6!A QE7!6d V!N<63 KH¬QEd!dD K6N7 d K6NEG

Vm = [αM0,k+1.pm0,k+1]y

Q/¡3QE D E([αM0,k+1.p

m0,k+1]y[α

M0,k+1.p

m′

0,k+1]y|(PN,mtk

)1≤m≤M)

=E([αM0,k+1.p

m0,k+1]y|(PN,m

tk)1≤m≤M

)E([αM0,k+1.p

m′

0,k+1]y|(PN,mtk

)1≤m≤M)

m 6= m′ !DE ! FE D!

αM0,k+1

MN"! ! ! //KHEJ (PN,m

tk)1≤m≤M,0≤k≤N

±YMN 3! f

! !-QE!ZV!VE PQE !PE K:FE 6!D KLN PE E $ ª" DMV) =ELDEJ3! @E EN=KL ¡ K6 KEJ¬ N E !=DE

f! !QE!

Z

¸ :MN'# LEVD Yt = y(t,Wt) = E

(φ(WT )

∣∣Ft)

T > 0

) W

K6K6NL o EJ!QE!!@! KL 1

aD·

Page 115: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³K6E " <³3KLEMNMV

(Y, Z)3FEH -!

Yt = φ(WT ) −∫ T

t

ZsdWs.

¥bK6 Q¬J EJ /EMk = E

(· |(Wm

tk)1≤m≤M

) ± εk

!N3 !FEHQEE-3! Q¡DK6KLHG

εk = E1

M

M∑

m=1

|yN,M (tk,Wmtk

) − y(tk,Wmtk

)|2

yN,M(tk, ·)

! Q¶DK6KL[αMk .pk]y(·)

pk(·)

QE!aD!¡D£E!QEJKk

^6D· apmk = pk(W

mtk

)

αMk5!6KL!KL!cDE 3G

infα

1

M

M∑

m=1

|yN,M (tk+1,Wmtk+1

) − α.pk(Wmtk

)|2

3 [ψ]y(·) = −‖φ‖∞ ∨ (ψ(·) ∧ ‖φ‖∞)

%aDψ

! ¯3E ¬K6"G8hf<`cf .£[ ,C[afW A A

L $ # &-$ +-

Kk(ω), +-$'& , + + "! "&- # $ , %$"+-$

tk "+-%,%,

M×Kk $ , , &-$ $ $ $ +-

(pk(Wmtk

)∗)1≤m≤M k = N−1

$ +

εN−1 ≤ infα

E|y(tN−1,WtN−1) − α.pN−1(WtN−1

)|2 +‖φ‖2

∞E(KN−1(ω))

M

' k ≤ N − 2

εk ≤ (1 + h)εk+1 + infα

E|y(tk,Wtk) − α.pk(Wtk)|2 + C‖φ‖2

∞E(Kk(ω)))

Mh.

,'SVegUh ¶S%ACAF $ ! +- + *! , + + + $ $ +- $ +'!

h−1 +-$ , #

C ‖φ‖2∞E(Kk(ω)))Mh

' + ' +- C ‖φ‖2

∞E(Kk(ω)))M

# $ + +- %, $ #$ + %, +-% +- ! $ +-$, !)+ $ #$ + %, +-% , #

hβ +

Cy(R)2 exp(− Mhβ+1

144Cy(R)2)E

(

N2

(h

β+12

3√

2, [P0,k+1]y−yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M)) +#

+- +-+- #$' +-$ # , + $ ‖·‖k+1,M

#$ ‖·‖k+1,M # $"!# + #$

Page 116: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @° »°7¯ ¸ ´ ¨¡¯ ¸¥b

k = N−1L3! D·KLNd!QEJdD£EJ!3!6 KLO N¥b

k ≤ N − 2V!

βMk/!5G

βMk = arg infβ

1

M

M∑

m=1

|y(tk+1,Wmtk+1

) − β.pk(Wmtk

)|2.

OP !5# E !!βMk

@E/G

EMk

1

M

M∑

m=1

|yN,M(tk,Wmtk

) − y(tk,Wmtk

)|2

≤EMk

1

M

M∑

m=1

|αMk .pmk − y(tk,Wmtk

)|2

=EMk

1

M

M∑

m=1

|αMk .pmk − EMk (βMk ).pmk |2 + EM

k

1

M

M∑

m=1

|EMk (βMk ).pmk − y(tk,W

mtk

)|2

= EFEH!¡ $E MNEMk (βMk )

3FEH -!

infα

1

M

M∑

m=1

|y(tk,Wmtk

) − α.pmk |2.

±Y!KLPKLP!-K6KHP!! PE !E)KLE PMNT1,k,M

¯E£! N!=E QEJ ¬ EMk

1M

∑Mm=1 |αMk .pmk − EM

k (βMk ).pMk |2 !εk+1

¸ MN

EMk

1

M

M∑

m=1

|αMk .pmk − EMk (βMk ).pmk |2 =EM

k

1

M

M∑

m=1

|EMk (αMk ).pmk − EM

k (βMk ).pmk |2

+ EMk

1

M

M∑

m=1

|αMk − EMk (αMk ).pmk |2

' E MN@E @!®DNED· !%FE D· ((pmk )∗

)

1≤m≤MK6LK6®! K6KH®!®! ¬¥ E'DN DK6K6® Eg! ! K6 DN 3!K6KL!®K6KH!)! <%)P 3E\!# EVFEJ !K6¡¤º!FE QEJ!6FE-KLEJD

EMk

(V − EM

k (V )V − EMk (V )∗

) V

D·L!RM !=DV!

([αMk+1.pmk+1]y)1≤m≤M

d¨"c!QE6-D£E6 o /! f

! !@QE!Z=K6K6 D5G

E1

M

M∑

m=1

|αMk .pmk − EMk (βMk ).pmk |2 ≤(1 + h)E

1

M

M∑

m=1

|αMk .pmk − βMk .pmk |2

+ (1 + h−1)E1

M

M∑

m=1

|βMk − EMk (βMk ).pmk |2.

Page 117: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³±¶E !DNED !FEgD· (

(pmk )∗)

1≤m≤MK6-! KLE 'K6 K6¡!@K6KH"!¡! QE

(1 + h)E1

M

M∑

m=1

|[αMk+1.pmk+1]y − y(tk+1,W

mtk+1

)|2 = (1 + h)εk+1.

¥bd 3D!K6NL dKK6EJK6dMNT3,k,M

ª DTK6! E¤ ¬ E !¡E: ¤OP 5MN /!E6-D£EJL o /! 5=! !E6!

Z! · EJ !5 EJH KL EJ !D! K6-´´ ®E5 )!5D KLN: KLNE ­PH !!"MV: ®¡EMN# E K6! D "E®! "!D·'E6!QELDE Ed

fH! !'!

Z)EJ) !L KLNEc E MN oE5MN/ aEJD 5KLE NQE ¥ EJ5KLEMN-!KL$J ! KL K6Nb¬Ed! FE3! KLEJ N ! MNDFEHJE5KLED $

#

¨ JE @D"MV# EVDE $EKLN/= D K6N= OP#¬5MN |∆Wml,k| + |∆Wm

l,k| ≤ Cw(R)√h Y³KLEMV"MN:# "!:VD£EJ EJ %

aD· /!R

"E "!¡E:KK6 ^E =MNE: -´´ ¥bPPE P ¡PQEJT\" KL EJN!H P E 3H $E ! a´´ D'MV¬5 NV!EJN)!: D K6H H KL¤ EJ$ <­E DE @%E MNE ®EJ K6"E K6 E <EE yN,R,Mk (·) = [αM0,k.p0,k]y(·)

zN,R,Ml,k (·) = [αMl,k.pl,k]z(·)

(αMl,k)0≤l≤q

3! Q¶E¡G

infαl

1

M

M∑

m=1

|yN,R,Mk+1 (PN,mtk+1

)∆Wm

l,k

h− αl.p

ml,k|2,

infα0

1

M

M∑

m=1

|yN,R,Mk+1 (PN,mtk+1

) + hf(tk, XN,mtk

, yN,R,Mk+1 (PN,mtk+1

), zN,R,Ml,k (PN,mtk

)) − α0.pm0,k|2.a´´ /±¶E®!T D-ECD ^´´ ¤ ´´ E5!D'MN-# E%K6ED

(PN,mtk+1

,∆Wmk )1≤m≤M

QE(PN,m

tk+1,∆Wm

k )1≤m≤M ¨=JE:D ¤º!MN!@ K6N3QE3 3=JE5@QE ¤D E 3FEL @MV<) ! $EJ a´´ M ¤º´´ !QEJD£EP!# E4K6ECD 3E E "G

Page 118: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @° »°7¯ ¸ ´ ¨¡¯ ¸8hf<`cf .£[ ,C[afW A A

O ' k ≤ N − 2

Eh

M

M∑

m=1

|αMl,k.pml,k − zN,Rl,k (PN,mtk

)|2

≤CCy(R)2

ME(Kl,k(ω)) + inf

αE|√hzN,Rl,k (PN

tk) − α.pl,k(P

Ntk

)|2 + Chβ

+ CE1

M

M∑

m=1

|[αM0,k+1.pm0,k+1]y − yN,Rk+1 (PN,m

tk+1)|2

− CE1

M

M∑

m=1

|E[αM0,k+1.pm0,k+1]y − yN,Rk+1 (PN,m

tk+1)|2

+ CE

(

N2

( hβ2

18Cw(R)2Kl,k(ω), [P0,k+1]y, (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M)e− Mhβ

36Kl,k(ω)Cy(R)2Cw(R)2

)

PN,mtk+1

= Tk(PN,mtk

,∆Wmk,k+1)

+ !(∆Wm

k,k+1)1≤m≤M'$ ! (∆Wm

k,k+1)1≤m≤M+ ! , + $ "+ $E(·) = E(·|(∆Wm

k,k+1)1≤m≤M,0≤k≤N−1)

,'SVegUh ¶S%ACA $ ! $ + $ +- # #, #, +-, & +-$ # ', + $ %$ !#

#$ $ #$ + %, +-% '$+-, & # $ # , #)+- ' , + "&-# $ , %$"+-$tk (PN,m

tk+1)1≤m≤M

, +- (∆Wm

k,k+1)1≤m≤M *!#$ # +# , #'$ %$ "& +-,

+ ! ! +- ! +-$1^´´ J ! ' $ %, ' + # , + $ $ + & + $ , ## ' +-$ , +-, & + # , +# +- $ #

hβ, ##$ # # +-$ , %$ "& +-, !# "+- #$ + #$ +-$ , + )+ $ , $ #$+- +-$ , +-, & + *! # ', + $ %$ !##$ $ #$ ,#$ +-%

³KLEMNMN

zN,Rl,k (PN,mtk

) = E(yN,Rk+1 (PN,mtk+1

)∆Wm

l,k

h)

(∆Wm

l,k)1≤m≤M D ! !QE !

(∆Wml,k)1≤m≤M

· E(·) = E

(· |(∆Wm

k,k+1)1≤m≤M,0≤k≤N−1

) ¬OP# NV!EJN'%D·βMl,k

!

infα

1

M

M∑

m=1

|yN,Rk+1 (PN,mtk+1

)∆Wm

l,k

h− α.pml,k|2

KE

Page 119: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³TVNG

E1

M

M∑

m=1

|αMl,k.pml,k − zN,Rl,k (PN,mtk

)|2 =E1

M

M∑

m=1

|αMl,k.pml,k − E(βMl,k).pml,k|2

+ E1

M

M∑

m=1

|E(βMl,k).pml,k − zN,Rl,k (PN,m

tk)|2.

±Y!KLLKLL!K6KHL!! LE LECD' K K6)EKLNMV'T4,k,l,M

Q¥b EJKL @E-G1

M

M∑

m=1

|αMl,k.pml,k − E(βMl,k).pml,k|2 ≤

2

M

M∑

m=1

|αMl,k.pml,k − αMl,k.pml,k|2

+2

M

M∑

m=1

|αMl,k.pml,k − E(βMl,k).pml,k|2

αMl,k

E:-!

infα

1

M

M∑

m=1

|[αM0,k+1.pm0,k+1]y

∆Wml,k

h− α.pml,k|2.

a´´ ±Y¡!K6"!¡# E 4 D ¤º!3¡E DK6K6" ¡G

E1

M

M∑

m=1

|αMl,k.pml,k − E(βMl,k).pml,k|2

=E1

M

M∑

m=1

|E(αMl,k).pml,k − E(βMl,k).p

ml,k|2 + E

1

M

M∑

m=1

|αMl,k.pml,k − E(αMl,k).pml,k|2

≤E1

M

M∑

m=1

|E([αM0,k+1.pm0,k+1]y − yN,Rk+1 (PN,m

tk+1))

∆Wml,k

h|2

+ E1

M

M∑

m=1

|αMl,k.pml,k − E(αMl,k).pml,k|2

≤h−1E1

M

M∑

m=1

|[αM0,k+1.pm0,k+1]y − yN,Rk+1 (PN,m

tk+1)|2

− h−1E1

M

M∑

m=1

|E[αM0,k+1.pm0,k+1]y − yN,Rk+1 (PN,m

tk+1)|2

+ E1

M

M∑

m=1

|αMl,k.pml,k − E(αMl,k).pml,k|2.

Page 120: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @° »°7¯ ¸ ´ ¨¡¯ ¸±YP! dK6EJ 3DK6KL

T5,k,l,M

D£EαM0,k+1

¬ ! T!QEJN¬!(∆Wm

k,k+1)m·¬E- · EP5 $EJ H!Lª¬ED ¤ ¸ D ¬EJ L!5E-KK6 ^E MV6 ^´´ J DKLK6: K6¡ ¡! K6"K6! KLKH:!6! :!5FE-! 6 ¤ $E ³3¡K6 1

M

∑Mm=1 |αMl,k.pml,k − αMl,k.p

ml,k|2

<°c=DK6QEEN MNQEJ a´´ / a´´ -KLE EJ MN"¬DV6DNαMl,k

αMl,k

N VD $ <ª "MN ¬K6 7 V!N$VD ¬MVP# L d! ! EJ ^EJ Nd K6KLN"! !6T !=/!pl,k

¸ E 6!* E a¡EO MNPMl,k = Id

QO1

M

M∑

m=1

|αMl,k.pml,k − αMl,k.pml,k|2 =(αMl,k − αMl,k).P

Ml,k (α

Ml,k − αMl,k)

=‖αMl,k − αMl,k‖2.

¨ β

¬MN1 < β ≤ 2

·5E!'FE%J EJPMk = P(·|(PN,m

tk)1≤m≤M )

¸ # NVPMl,k = Id

@E

PMk

(

‖αMl,k − αMl,k‖2 ≥ hβ−1

)

=PMk

(Kl,k(ω)∑

i=1

∣∣

1

M

M∑

m=1

pml,k,i[αM0,k+1.pm0,k+1]y

∆Wml,k

h− [αM0,k+1.p

m0,k+1]y

∆Wml,k

h∣∣2 ≥ hβ−1

)

≤Kl,k(ω)∑

i=1

PMk

(∣∣

1

M

M∑

m=1

pml,k,i[αM0,k+1.pm0,k+1]y

∆Wml,k

h− [αM0,k+1.p

m0,k+1]y

∆Wml,k

h∣∣2 ≥ hβ−1

Kl,k(ω)

)

≤Kl,k(ω)∑

i=1

PMk

(

∃ψ ∈ [P0,k+1]y :∣∣

1

M

M∑

m=1

pml,k,iψ(PN,mtk+1

)∆Wm

l,k

h− ψ(PN,m

tk+1)∆Wm

l,k

h∣∣2 ≥ hβ−1

Kl,k(ω)

)

=

Kl,k(ω)∑

i=1

PMk

(

∃ψ ∈ [P0,k+1]y :∣∣

1

M

M∑

m=1

pml,k,iUmψ(PN,mtk+1

)∆Wm

l,k

h− ψ(PN,m

tk+1)∆Wm

l,k

h∣∣ ≥

hβ−1

Kl,k(ω)

)

(Um)

!JEFE3E EJ# # ! ^·P! !ENP!P PEJJEJFEJE EJ7 K: EFEJN1

−1E£DPQE 1

2

V´ V! EdPD¦¤ G ![P0,k+1]y

E dψ ∈ [P0,k+1]y

VQ gj ∈ G V PQE

Cy(R) "MN1

2M

M∑

m=1

|ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 + |ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 ≤ hβ

18Cw(R)2Kl,k(ω)

ET ®MNCw(R)

/"MN |∆Wml,k| + |∆Wm

l,k| ≤ Cw(R)√h ¯-MV

D£EJ! QEc! G N2

(h

β2√

18Cw(R)2Kl,k(ω), [P0,k+1]y, (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M) ¶O

M

Page 121: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

ψ ∈ [P0,k+1]y·gj

<EJN# E 4 D ¤º! T¨@E

∣∣

1

M

M∑

m=1

pml,k,iUmψ(PN,mtk+1

)∆Wm

l,k

h− ψ(PN,m

tk+1)∆Wm

l,k

h∣∣

≤∣∣

1

M

M∑

m=1

pml,k,iUmψ(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣

+∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣

+∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− ψ(PN,m

tk+1)∆Wm

l,k

h∣∣

¨@E

∣∣

1

M

M∑

m=1

pml,k,iUmψ(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣2

=1

M

M∑

m=1

|ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2|∆Wm

l,k

h|2

≤2Cw(R)2

2hM

M∑

m=1

ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2 + |ψ(PN,mtk+1

) − gj(PN,mtk+1

)|2

≤ hβ−1

9Kl,k(ω)

EJNdMN 1M

∑Mm=1 |pml,k,i|2 = 1

QEV <ª KLK63FEKK63 E JEJFE K6 ∣

∣ 1M

∑Mm=1 p

ml,k,iUmgj(PN,m

tk+1)

∆Wml,k

h− ψ(PN,m

tk+1)

∆Wml,k

h∣∣Q/ ¡G

PMk

(

∃ψ ∈ [P0,k+1]y :∣∣

1

M

M∑

m=1

pml,k,iUmψ(PN,mtk+1

)∆Wm

l,k

h− ψ(PN,m

tk+1)∆Wm

l,k

h∣∣ ≥

hβ−1

Kl,k(ω)

)

≤PMk

(

∃j :∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣ ≥ 1

3

hβ−1

Kl,k(ω)

)

.

Page 122: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«OP¥c´ ³° @° »°7¯ ¸ ´ ¨¡¯ ¸¥bvYDc!vK6¡D! ¡YQEvEJ !

(∆Wmk,k+1,∆W

mk,k+1)1≤m≤M

¨=PMk+1

D "QE4 ´ vV N¡G

PMk+1

(

∃j :∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣ ≥ 1

3

hβ−1

Kl,k(ω)

)

≤N2

(h

β2

18Cw(R)2Kl,k(ω), [P0,k+1]y, (P

N,mtk+1

)1≤m≤M , (PN,mtk+1

)1≤m≤M

)

maxj

PMk+1

(∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣ ≥ 1

3

hβ−1

Kl,k(ω)

)

¯JXm = pml,k,iUmgj(PN,m

tk+1)

∆Wml,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h E

EMk+1(Xm) = 0

· |Xm| ≤2Cy(R)Cw(R)√

h|pml,k,i|

Q°c%E MNQEN# $EJ !«V !=N¡G

PMk+1

(∣∣

1

M

M∑

m=1

pml,k,iUmgj(PN,mtk+1

)∆Wm

l,k

h− gj(P

N,mtk+1

)∆Wm

l,k

h∣∣ ≥ 1

3

hβ−1

Kl,k(ω)

)

≤2 exp

(

− Mhβ

36Kl,k(ω)Cy(R)2Cw(R)2 1M

∑Mm=1 |pml,k,i|2

)

=2 exp

(

− Mhβ

36Kl,k(ω)Cy(R)2Cw(R)2

)

E"# NV 1M

∑Mm=1 |pml,k,i|2 = 1

PMl,k = Id

¶¨"E-!D6 !-E!:FE K6 Nαl,k

·αl,k

E3 ¬KLKLN! !¡ (!LH!pl,k

¤

Page 123: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´ QO±\µ¨¡³´ «I@° I®¨"­P´ ´¦° ¥d¨"©³u±¶° ¸ °7­ ¸ ³

Page 124: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

l@mxpY-pr | j.sl-t@pv|

| |gn t=|wk-pvmwk

Page 125: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 126: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

pt/|-'| q| ~ mw |l=x|wk~| ksywk v| ~syql@p1|l YpYw sxpvl=|

¯3ECN@!QEJ3E:E"´EHV max

0≤k≤NE|Ytk − Y N

tk|2 ≤ C

√h

:# E KLE !Y

E5E!= °7­ ¸ ³! D=! Q /QE^´ C¤ ´ ­E")DE )! f

" £E (y, z)

b" )! P)V :EJ!) EJKLEJ %!Y

QEY N v¯¡E P! K6N$ ENP ! ! · !QE ²¡«P µI% Y ²¡«P¥7 QMN¡# /NEHV JE5G

|Yt0 − Y Nt0| ≤ Ch.

ª H E!QEP# P!PE£JE%! ³PI M b ¥7 M M ® PEPD!N=D£EJ!¡ EYKLE¬ N3QE!"V !¡DND

¥c D! E!/NVMN EJ ECE G> v`7f-, Z@ .S%A ACA

$ d = q = 1 , $ ! $ b(t, x) σ(t, x)

$ +-$ , $ $ # #$ , $ $ C∞

b

+-$ , + ' $ $ , + $ ! $ φ(x) ' $ , + ! $ - $ #%$+-, , + +- ,! $ ,

|φ(x)| ≤ C exp(C|x|).

$ $ $ σ '$ #$ #,%, + #$ |σ(t, x)| ≥ σ0 > 0 ∀(t, x)

Page 127: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸¯E E K6¡MV6# °7­ ¸ ³ ^´ /"E- ! ¡ E )

(y, z)D! ¡KL QE c! !QE MNK6 !EE¶K6QEJ

XT

°7)QEJ D $£E ¬D! 3 °d­ ¸ ³ EN6G

−dYt = −(θ(t,Xt)Zt + rYt)dt− ZtdWt,

YT = φ(XT ),

θ(t, x) = b(t,x)−r

σ(t,x)

¬!PDFEJC∞b

¬ VP´´´ ­PPKK63MV QE D !$JKH

XN 7D KLE3! °c \!X

·b)EVD °7­ ¸ ³ !K6\D VQEY NtN

=φ(XNT ),

^´´´ /Y Ntk

=Etk(YNtk+1

) − hθ(tk, XNtk

)ZNtk− rhY N

tk,

^´´´ hZN

tk=Etk(Y

Ntk+1

∆Wk).^´´´

#

(+*& #2 /:<!¯3E ¬!D¡TD3!¡! KLN K6¡ JEJNG()ZJNf<h @VegS@ACA A

, $ +

|Yt0 − Y Nt0| ≤ Ch.

­E3D¡D QE = EHFE aD·η : [0, T ] → t0, · · · , tN

! ¡QEJ

η(t) =N−1∑

k=0

tk1[tk,tk+1[(t) + T1T (t).

c a´´´ ¤ ´´´ Q=¡GY Ntk

= Etk

(Y Ntk+1

1 − θ(tk, XNtk

)∆Wk

1 + rh

)

D¡MNY= Ek

!6GY Nt0

= E( 1

(1 + rh)Nφ(XN

T )ξNT)

ξNT =

N−1∏

k=0

(1 − θ(tk, XNtk

)∆Wk).

Page 128: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³° ´ ° ¸¸ ° ­° ª¨"¯ °7³µ¡°7¯ª¬° ­³´ "°7³u±Y´¦¯P°cOP´¦³3°± N · ! D

Y Nt0

cD aK6 cMN 6E MVEN EP O JH ¬GYt = E

(φ(XT )ΓtT |Ft

)

dΓts = −Γts(rds+ θ(s,Xs)dWs) ,Γtt = 1.

°c@ JEJND!°7­ ¸ <=Y0 = E

(exp(−rT )φ(XT )ξT

)

ξT = exp(−∫ T

0

θ(t,Xt)dWt −1

2

∫ T

0

|θ(t,Xt)|2dt).

¨DQEJ E =!@EEJ MN ¤ !EC6MVY0

$ KLDKLK6®# EJD ! QE£ EDQE /FEQEJ % MN ¤ Q

! QQEdQdP

∣∣FT

= ξT OP |Yt0 − Y N

t0| KLE QEFEHKLK6"!" ¬K6¡G

∣∣E(exp(−rT )φ(XT )ξT

)− E

(exp(−rT )φ(XN

T )ξT)∣∣

+∣∣E(exp(−rT )φ(XN

T )ξT)− E

(exp(−rT )φ(XN

T )ξNT)∣∣

+∣∣E(exp(−rT )φ(XN

T )ξNT)− E

( 1

(1 + rh)Nφ(XN

T )ξNT)∣∣.

¯3E ¬3D QEJD@!¡D KL$ (hf[ .£[B@eS',CSNhegS

ª ¡KL"¡%^ED K6=KEJMNQE3MNexp(−rT ) − 1

(1 + rh)N= exp(−rT ) − exp(−N log(1 + rh))

= exp(−rT )(1 − exp[rT − T

h(rh− r2h2

2+ o(h2))]

)

=Ch+ o(h).

±Y! @EK6 D N!3 EJNKL !FE $ +* (+*#&' 2 #*/7 # (+-'!+*¯3E! E! ¯E M YQED

Dk,p !=DEDv!I%E E£V

E

Page 129: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸07SegegS@A ACA

$ ! $# # '$ -#$# $ d (Xt)t≥0

$ +- ! !# #$

C∞ # #$ (XNt )t≥0

$ ! + , #

Xt = x+

∫ t

0

b(s,Xs)ds+

q∑

j=1

∫ t

0

σj(s,Xs)dWjs ,

XNt = x+

∫ t

0

b(η(s), XNη(s))ds+

q∑

j=1

∫ t

0

σj(η(s), XNη(s))dW

js .

, 1 ≤ k ≤ d

Xk,t −XNk,t =

q∑

i,j=0

cX,0i,j,k(t)

∫ t

0

cX,1i,j,k(s)∫ s

η(s)

cX,2i,j,k(u)dWiudW j

s

+ *!, + ! $ #$ $ dW 0t = dt

! + +# (cX,i1i,j,k (t))t≥0 : 0 ≤ i, j ≤q, 1 ≤ k ≤ d, 0 ≤ i1 ≤ 2 + +-+-$ supN,t ‖cX,i1i,j,k (t)‖k′,p <∞ ' k′, p ≥ 1

¯3E $EJKLN =!¡FE aKH ! EJ-QEQE JE5 §¬¯ M G8hf<`cf .£[ ,C[afW A ACA

F ∈ (D∞)m

#, + + "! +-,%, + %$γF

%$ # # , + + γ−1

F ∈ ∩p≥1Lp φ ∈ C∞

b (Rm)G ∈ D∞ , (det γF )−1 ∈ D∞

' ', %$ -"! α %, '$ +- +, +-,)+ % Hα(F,G) ∈ D∞ #,%, E(G(∂αxφ)(F )

)= E

(φ(F )Hα(F,G)

).

, p > 1 , %$ -"! α %, '$ ! $ +-$ C(p, α) #$ # $+ #,

n1, n2 %$ -"!

k, d, d′, b, b′ #$ +-$ "& +-,#$

pα #,

‖Hα(F,G)‖p ≤ C(p, α)(‖γ−1

F ‖n1k ‖F‖n2

d,b‖G‖d′,b′).

±Y7 K6d JEJN KLQE !QE ¥c M CMVN KL7 aK67 D ! °7­ ¸ £EJVE5 $EJKLN¡G()ZJNf<h @VegS@ACA A

F H '$ +- %$'& +-,

Z'$ +- %$'& +-, ! $ %$ #,%,

Z0 = 0 , , + , $ EH(Z) , + $

Xt = Ht +

∫ t

0

Xs−dZs

$ $ +-

EH(Z)t = E(Z)tH0 +

∫ t

0+

E(Z)−1s d(Hs − [H,Z]s)

E(Z)t = exp(Zt − 12[Z,Z]t)

Page 130: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³° ´ ° ¸¸ ° ­° ª¨"¯ °7³µ¡°7¯ª¬° ­³´ "°7³u±Y´¦¯P°cOP´¦³3°°c)QQ3E£=!=K6KL EN¡G07SegegS@A ACA

F q, p > 1 $ +

supt

‖Xt‖q,p + supt

‖ξt‖q,p + supt,N

‖XNt ‖q,p + sup

k,N‖ξNtk‖q,p <∞

+ *!(ξNt )t

$ +-

ξNt = 1 −∫ t

0

ξNη(s)θ(η(s), XNη(s))dWs.

±Y: EJ"P K6PK6"! D ^EJD K6N! ^EJ MNb, σ

θL ! ! L $ d¥b5-!5KL c E /N =!

6D ( XN

ξN

) DK6KL6LD KLE®! °c)!# °7­ ¸ ( X

ξ

) ·:E MN)E! D·KLN'!- EJDFEMN' ! !%!-D L!QE

Dq,p ¬I%EJ DV6D N3!¡ °7­ ¸ ( X

ξ

) ¡3QE(!:! 3 !:DE!ξ ¡

QE D ! ¡TD· = EN¬FEH !¡!θ \3! E!QK6NMN

supk≤N,N ‖ξNtk‖p <∞ p ≥ 1

7@E:! Q /!ξNt

QE51 ≤ i ≤ N

GξNti = ξNti−1

(1 − θ(ti−1, XNti−1

)∆Wi−1).^´´´

OP =Np ≥ 1

G

E|ξNti |2p =E

(

|ξNti−1|2pEti−1

2p∑

k=0

Ck2p(−1)2p−kθ(ti−1, X

Nti−1

)k∆W ki−1

)

=E

(

|ξNti−1|2pEti−1

p∑

k=0

C2k2p (−1)2(p−k)θ(ti−1, X

Nti−1

)2k∆W 2ki−1

)

≤(1 + Cph)E(|ξNti−1|2p)

h

LEK6KLNT· / E¬E)!" $EJ PE) !"!θ Q¨=! ! 3E ^ED KLNMN

supk≤N,N ‖ξNtk‖p <∞ °c dD! u

MNtk−1 < u ≤ tk

d°co EJN6MVaD ' $ -!EDDK6:$ D QEMNξNti ∈ D∞ °c!T NFE'a´´´ H

i ≥ k + 1G

DuξNti

= DuξNti−1

−Du[ξNti−1

θ(ti−1, XNti−1

)]∆Wi−1.

M

Page 131: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸°c@ E=

i ≥ k + 1G

DuξNti

= DuξNtk−

i−1∑

j=k

Du[ξNtjθ(tj, X

Ntj

)]∆Wj.

°c E MNQEJN $EJ ! § N! ¤ ­ECV ¤ µ"! ! FE KLEE ∑

j Du[ξNtjθ(tj, ξ

Ntj

)]∆WjTVG

E|DuξNti|p ≤CpE|Duξ

Ntk|p + CpE

∣∣

i−1∑

j=k

Du[ξNtjθ(tj, X

Ntj

)]∆Wj

∣∣p

≤CpE|DuξNtk|p + CpE

∣∣h

i−1∑

j=k

|Du[ξNtjθ(tj, X

Ntj

)]|2∣∣

p2

≤CpE|DuξNtk|p + CphE

i−1∑

j=k

|Du[ξNtjθ(tj, X

Ntj

)]|p

≤Cp(1 + E|DuξNtk|p) + Cph

i−1∑

j=k+1

E|DuξNtj|p

D£EJθ

‖ξNtj ‖q · ‖XNtj‖1,q

N$q ≥ 1

-Q 3aK6 K63N

j °c%EJ MNQEN "KLK6!µ"¬E !D·Q=¡G

maxk≤i≤N

E|DuξNti|p ≤ Cp(1 + E|Duξ

Ntk|p).

¥ EE GDuξ

Ntk

= −ξNtk−1θ(tk−1, X

Ntk−1

)

! ==! ! maxk≤i≤N

E|DuξNti|p ≤ Cp(1 + E|ξNtk−1

|p).

±Y! 3! !¡ "EJ N!¡FEHK KL"^E ¤

`= )+& $-& "'(+ ¯!: D VFEH! KLEJ -!- KL"´´´ ·3!EK63!E:K6KL 8hSVe[aSNh*,£SVhegS

C

Page 132: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³° ´ ° ¸¸ ° ­° ª¨"¯ °7³µ¡°7¯ª¬° ­³´ "°7³u±Y´¦¯P°cOP´¦³3°­P QFEHKL ||| · |||2

E¡G|||φ|||2 =

E|φ(XT )|2 +√

E|φ(XNT )|2 +

(E

∫ 1

0

|φ(XN,λT )|2dλ

) 12

XN,λt = λXt + (1 − λ)XN

t

Y¯3EJP¡! E! EJ¡ DMNφ !DFE

C∞ !5 DK6QED C∞

0

-!5D! /! KLMN∣∣E(exp(−rT )φ(XT )ξT

)− E

(exp(−rT )φ(XN

T )ξT)∣∣ ≤ C|||φ|||2hECD

φ!¡DFEJ

C∞0

Q°c= ·DE!φ

KLE @! ! /D ! EN" ! aDφm

!EC∞

0

MV!5φ

cFEKL ||| · |||2 Nª 3V3TD· =3 D¡ DKLK6

E(exp(−rT )φ(XT )ξT

)− E

(exp(−rT )φ(XN

T )ξT)

=

∫ 1

0

E(exp(−rT )ξT∂xφ(XN,λ

T )(XT −XNT ))dλ.

°c@ EJN "KLK6´´´ Q=NE(exp(−rT )ξT∂xφ(XN,λ

T )(XT −XNT ))

=1∑

i,j=0

E(exp(−rT )ξT∂xφ(XN,λ

T )cX,0i,j (T )

∫ T

0

cX,1i,j (s)(

∫ s

η(s)

cX,2i,j (u)dW iu)dW

js

).

¯E E ¬ PK6"D!QEJN!i = j = 1

°c= EJN¬E aKH P!"!QEJ ¯E M 3 KL¡! Q/ ¡G

E(exp(−rT )ξT∂xφ(XN,λ

T )cX,01,1 (T )

∫ T

0

cX,11,1 (s)(

∫ s

η(s)

cX,21,1 (u)dWu)dWs

)

=

∫ T

0

E(Dsexp(−rT )ξT∂xφ(XN,λ

T )cX,01,1 (T )cX,11,1 (s)(

∫ s

η(s)

cX,21,1 (u)dWu))ds

=

∫ T

0

∫ s

η(s)

E

(

Du

[Dsexp(−rT )ξT∂xφ(XN,λ

T )cX,01,1 (T )cX,11,1 (s)]cX,21,1 (u)

)

duds.

±Y@KLE(Du[Dsexp(−rT )ξT∂xφ(XN,λ

T )cX,01,1 (T )cX,11,1 (s)]cX,21,1 (u)) '% D %FE'aK66! K6K66!LK6

E(∂αxφ(XN,λ

T )G) MN HKE ¡ EFE )´´´ c# $EJ !3ª¬ED N¤ ¸ D ¬E 3E

C‖φ(XN,λT )‖2

³KLEMNcMN ¶FE ´´´ D£EJ(XN,λ

t )t )VD ! ´º aK6 K6\ MN

²¡«P¥7 Q´ DFEH!K6d = 1

3DDE"·3 V ¡!3! 3DKLK6"!QEJ µI% # OP =NQEKLNMNE(exp(−rT )ξT∂xφ(XN,λ

T )cX,01,1 (T )

∫ T

0

cX,11,1 (s)(

∫ s

η(s)

cX,21,1 (u)dWu)dWs

)≤ C‖φ(XN,λ

T )‖2h.

C

Page 133: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸±Y3EK6¡E !¡KK6" QQE K6DD !

∣∣E(exp(−rT )φ(XT )ξT

)− E

(exp(−rT )φ(XN

T )ξT)∣∣ ≤ C|||φ|||2h.#-SN¶[B@VegS,CSNhegS

±¶¡! K6"KL∣∣E(exp(−rT )φ(XN

T )ξT)− E

(exp(−rT )φ(XN

T )ξNT)∣∣

K6 # E KLEJ )! ! ¬!³3E!ξT

QE ¬EE ¬EJ EJ d! ED1

KEJJEH!'!)JEJ) $EJ ξNT

ª K6K6L) LK6¡K6\3MNφ

3!QEC∞

0

!5D! -!¡KLMN∣∣E(exp(−rT )φ(XN

T )ξT)− E

(exp(−rT )φ(XN

T )ξNT)∣∣ ≤ Ch|||φ|||2.¯"E KLPFEL! D

ξT − ξNT!' E!!® KL)´´´ v³KLEMNP! E!=MV

dξt = −ξtθ(t,Xt)dWt3MNξNT

EHEKL QE !¡ MNQEJ /! ´º ξNt = 1 −

∫ t

0

ξNη(s)θ(η(s), XNη(s))dWs.

¨=!D¡ KLE:! D¡DK6K6ξt − ξNt =

∫ t

0

ξNη(s)θ(η(s), XNη(s))dWs −

∫ t

0

ξsθ(s,Xs)dWs

= −∫ t

0

(ξs − ξNs )θ(s,Xs)dWs +

∫ t

0

[ξNη(s)θ(η(s), XNη(s)) − ξNs θ(s,Xs)]dWs

=E(α)t

∫ t

0

E(α)−1s εNs [dWs + θ(s,Xs)ds]

αt = −

∫ t

0θ(s,Xs)dWs

a vb KLc´´´ ¬YFE EJ E(·) TεNs = ξNη(s)θ(η(s), X

Nη(s))−

ξNs θ(s,Xs) ¨=PEJ D ¡KL"!

εNs3FE aK6:G

εNs =ξNη(s)θ(η(s), XNη(s)) − ξNs θ(s,Xs)

=ξNη(s)θ(η(s), XNη(s)) − ξNη(s)

[1 − θ(η(s), XN

η(s))(Ws −Wη(s))]θ(s,Xs)

=ξNη(s)[θ(η(s), XNη(s)) − θ(s,Xs)] + ξNη(s)θ(η(s), X

Nη(s))θ(s,Xs)(Ws −Wη(s))

=ξNη(s)[θ(η(s), XNη(s)) − θ(η(s), Xη(s))] + ξNη(s)[θ(η(s), Xη(s)) − θ(s,Xs)]+

ξNη(s)θ(η(s), XNη(s))θ(s,Xs)(Ws −Wη(s)).¨@JELEJE MV3E6KK6¡D MNMV) K6K6<DKH QE K6! Q)aK:"!!QE ¡·! N $EJ /QEQE !6D! =MN

ξNtk ∈C

Page 134: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«3O¥c´ ³° ´ ° ¸¸ ° ­° ª¨"¯ °7³µ¡°7¯ª¬° ­³´ "°7³u±Y´¦¯P°cOP´¦³3°

D∞ MNsupN,k ‖ξNtk‖k′,p < ∞

k′, p > 1 bª" )ED·K6)5 E!g±YK6K6´´´ Q´ v! D QEK6N

∣∣E(exp(−rT )φ(XN

T )ξT)− E

(exp(−rT )φ(XN

T )ξNT)∣∣ ≤Ch‖φ(XN

T )‖2

≤Ch|||φ|||2.

¤

C

Page 135: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸

C

Page 136: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

i vymxl@pt@kr | j mxql v| kst@pvmwkkpT l=|w1t=pv|Yv| 6t=m~ksx6t@p q| l t=l=mxyl%syq|l ~kpv|

#

¨=D !¡E:KK6"!VQEK6 MV¡X

Xi,t = Xi,0 +

∫ t

0

bi(s,Xs)ds+

∫ t

0

σi(s,Xs)dWs, 1 ≤ i ≤ d

3=)aD¡# NV 5G> v`7f-, Z@ .S%A ACA

F # $ , $ ! $ b

σ $ ! $

$ - # #$ +, + *! # + +-, $ '$ #$ $ $ $ '$ ! $ - $ '$ #,%, "!# '

σσ∗ σσ∗ ≥ ε0Id+ *!

ε0 > 0 °c' ·V¬E !3E£JE dE£D'VD¬! ´º )aK6 K6d MN EJ MNd E! I v ¯ ¬ $EKLNMNP JP´ P Q ­ED·QE QD QE# °d­ ¸ ³ a´ /¬PEJ! KE D D P=(Y, Z,K)

MNY Q5GYt = g(T,XT ) +

∫ T

t

f(s, Ss, Ys, Zs)ds−∫ T

t

ZsdWs +KT −Kt,

Yt ≥ g(t,Xt), t ≥ 0,

∫ T

0

(Yt − g(t,Xt))dKt = 0.^´´´ /

¯\b a ! °d²¡²¡¥ + M E N¶¶\QED !E\MN\D D b(Y, Z,K)

bEJ MN'TH!: EJH! D/:! D ) E K6

Page 137: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸# E3O N± N EJ QQEJD 3 FE" JE ¨L! 7D a !3JEJ

(Yt)t·

(Zt)t !FE¡ 5'EJD

S!E 3 a ª a $T K6T) MN:FELE!a P !ED

g(t,Xt) ª EDK6 E!D! EK6 DE 7 QE\ KL!QE3 "D£E! =EK6 D£EJ@EEJ

g(t, x) = (K − ex)+ ¯E c!QE VD E6D QE MV <dd KL!3KLV!Qc dE K6 3!QEE´3´´ DK6¡!¡E:DNE N"!FED $ ° E ! MV¬ D!5E K57 bD MN ! D!N !3!bV !)D D5P D MNH!QEPFELE´ T%JE P "FE6KLV!QD£E !-D6E K6H :DK6-!FEJD $ c± E!QEJ EJ! K6:E K6 N ECD) EJ !H¥c D£E!< K6 E <±YNVP ED

gNP JEJN

> v`7f-, Z@ .S%A ACAg(·, ·) , # #$ $ %$ -"! 1

2

#$ '$ #$ !'#$ , + +- +, + +-, ³KLEMN"MN:HK6 :! EJKL DE % ! KL

1vD:NV - Q · E3! MN ED

(K − ex)+ NYx ≥ log(K)

¨ ! Q gR

!5KEJ:EQE !φN,R

fR

!QE¡E-QEJ :´´ ¶±Y¡Cy(R)

·Cz(R)

)! QDK6K65 D !K6K6 b©3gDE Dc KL EJ *!=D 7!6FE- ´´ K63!3¬JE D KLN$ ± EJ K6 N3aKL E!:G→→→ ¨= E

k = NQE

yN,R,MN (·) = gR(tN , ·)

→→→ Oy% EP!)! D EJ tk(αMl,k)1≤l≤q

! QbDK6K6®K6 K6 3!%J¤ K6"!¡KL!¬D£E ¡Ginfαl

1

M

M∑

m=1

|yN,R,Mk+1 (XN,mtk+1

)∆Wm

l,k

h− αl.p

ml,k|2

ECDXN,mtk+1

KH !¬FEKKL ^E HMVPN,mtk+1

!QE EQE ´´ !PE b!XN,mtk

·c ¤D K6\ E E

∆Wmk

QEXN,mtk+1

= XN,mtk

+b(tk, XN,mtk

)h+σ(tk, XN,mtk

)∆Wmk

¨! Q ¡E

zN,R,Ml,k (·) EzN,R,Ml,k (·) = [αMl,k.pl,k]z(·)

¥cαM0,k

¡! Q DKLK6K6 K6 ¬!=K6"!K6 !DE G

infα0

1

M

M∑

m=1

|yN,R,Mk+1 (XN,mtk+1

) + hf(tk, XN,mtk

, yN,R,Mk+1 (XN,mtk+1

), zN,R,Ml,k (XN,mtk

)) − α0.pm0,k|2.

¨=! Q 3EyN,R,Mk (·) QEJ

yN,R,Mk (·) = max([αM0,k.p0,k]y(·), gR(tk, ·)

) C

Page 138: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP±\µ¨"³´ «I%°¥d¨"©P³±\° ¸ °7­ ¸ ³3³

→→→ ¨= MN !6# Et0

¨b b KE Y7KK6dE K6cEb E E aK KL7KLEMNMN FED· VQE´´ OP V# E K63MNQEK6¬K:FE "!)DT! ·¤D c!QEcFEQE ¬´´ 3!"DD<7MN !¡D QEJMV3 E*aDE KEJ yN,R,Mk! )E@ED

gR(tk, ·) ±YKK6"¦V! E K6¡E6! ¦!6 ·EQEJ V !QE §¬¥7¥7 § KLE E)!HD MV:! :!'MVEN ¤

QDEJ ¡:K6¡EDH!5D£E D !I@E E£ ¡:D! b­EJD!E D$ :DE¡! ! "! !EN!Z

QE"NV E 5E MN5E D·= 'D# ¯356 E K6N:! E5 V6! E K5H5 °7­ ¸ ³ & D 3!QEFEHQE "VKL MN"KEJE£DEE V"K6 MV ¯PE K6 FE MN ! DK6K6 N(yN,R,M (XN), zN,R,M (XN))

(Y N , ZN )

! D E:5KL7! ^´´´ /bMV3! D KEJNQEJN$

­PHKLE)EJQE ! ^´´ ¤ ´´ ®! Q H! D E(Y N , ZN )

! ^´´´ /!¡KLE ·E!5G→→→ ¨=

Y NtN

= g(tN , XNtN

)

→→→ O = ENtkQ/! Q

ZNtk

QE

hZNtk

= Etk(YNtk+1

∆Wk),^´´´

/! Y Ntk

YMN

Y Ntk

= Etk(YNtk+1

) + hEtkf(tk, XNtk, Y N

tk+1, ZN

tk),

^´´´ 3)Q@=

Y Ntk

= max(g(tk, X

Ntk

), Y Ntk

) ¨ ! ¡ E K6": VDE

(Y N,R, ZN,R) K6E £EJN

gf

EgR

·fR

O D£E%!@E =!®I%E !

XN E Y N,Rtk

= yN,Rk (XNtk

)Y N,Rtk

=

yN,Rk (XNtk

)ZN,Rtk

= zN,Rk (XNtk

)

­PHK K6)MN:E @´´ T" EJ :H PDKLK6=VDE E °7­ ¸ ³ & D

C9E

Page 139: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸8hf<`cf .£[ ,C[afW A ACA

F

max0≤k≤N

E|Y Ntk

− Y N,Rtk

|2 + hE

N−1∑

k=0

|ZNtk− ZN,R

tk|2

≤CN−1∑

k=0

E[1g(tk,XN

tk)6=gR(tk,X

Ntk

)|Y Ntk|2 + Cy(R)2

]+ CE(|g(tN , XN

tN) − gR(tN , X

NtN

)|2)

+ ChE

N−1∑

k=0

|f(tk, SNtk, Y N

tk+1, ZN

tk) − fR(tk, S

Ntk, Y N

tk+1, ZN

tk)|2.

E|Y Ntk

− Y N,Rtk

|2

=E[|g(tk, XN

tk) ∨ Y N

tk− gR(tk, X

Ntk

) ∨ Y N,Rtk

|21g(tk,XNtk

)=gR(tk,XNtk

)

]

+ E[|Y Ntk

− Y N,Rtk

|21g(tk,XNtk

)6=gR(tk,XNtk

)

]

≤E[|Y Ntk

− Y N,Rtk

|21g(tk,XNtk

)=gR(tk,XNtk

)

]+ 2E

[(|Y N

tk|2 + Cy(R)2)1g(tk,XN

tk)6=gR(tk,X

Ntk

)

]

≤E[|Y Ntk

− Y N,Rtk

|2]+ 2E

[(|Y N

tk|2 + Cy(R)2)1g(tk,XN

tk)6=gR(tk,X

Ntk

)

].

±Y:KLE[|Y Ntk

− Y N,Rtk

|2] E MN5H5!:FE'! K6EJ®:E ¡E!:FEKK6"KLEMNFEH '´´

¤

#

´ < ^ED !7MN3c KL´´ N´´ )´´ ´´ ¡Nd7JEE $ °7LT$ 63!¡KLEMNMN |a ∨ b− c ∨ b| ≤ |a− c| ·3!D5G1

M

M∑

m=1

|yN,Rk (XN,mtk

) − yN,R,Mk (XN,mtk

)|2

=1

M

M∑

m=1

|gR(tk, XN,mtk

) ∨ yN,Rk (XN,mtk

) − gR(tk, XN,mtk

) ∨ [αM0,k.p0,k]y(XN,mtk

)|2

≤ 1

M

M∑

m=1

|yN,Rk (XN,mtk

)) − [αM0,k.p0,k]y(XN,mtk

)|2

3 EQE " E "!QEFEHQE ´´ 3EFEJD¡!K6 C

Page 140: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP±\µ¨"³´ «I%°¥d¨"©P³±\° ¸ °7­ ¸ ³3³

´ " ! · E FE DNDg!(Y N , ZN )

(Y, Z)

¥\/DFE)¬ JEDK6QE(Y N , ZN )

!5# MVEJ / E!"!(Y N,1, ZN,1)

Q NV! !QEJ I v cG→→→ ¨= E

tNQEJ

Y N,1tN

= g(tN , XNtN

)

→→→ ¥c%t ∈ [tk, tk+1[

MNQEJ E!g & D g! (Y N,1, ZN,1)

GY N,1t = Y N,1

tk+1+

∫ tk+1

t

f(s,XNs , Y

N,1s , ZN,1

s )ds−∫ tk+1

t

ZN,1s dWs.

^´´´

→→→ ¨=! Q ¡t

!QE[tk, tk+1[

Y N,1t = Y N,1

t ∨ g(t,XNt )

→→→ ¨= MN !6FEH!QEJ

t0 = 0

8hf<`cf .£[ ,C[afW A ACA h + #$ $ +

max0≤k≤N

E(|Y Ntk

− Y N,1tk

|2) +N−1∑

k=0

E

∫ tk+1

tk

|ZN,1t − ZN

tk|2dt ≤ Ch+ CE

N−1∑

k=0

∫ tk+1

tk

|ZN,1s − Z

N

tk|2ds

hZN

tk= Etk

∫ tk+1

tkZN,1s ds

9°c EN |a ∨ b − c ∨ b| ≤ |a − c| ·5 EMNQEJNHDKLK6LH $EJ ! d)5!5QEJ ¬!3 MNQEJ a´´´ ¤ ´´´ Q=N3E K6N¡GE|Y N

tk− Y N,1

tk|2

≤E|Y Ntk

− Y N,1tk

|2

≤(1 + γh)E|Etk(YNtk+1

− Y N,1tk+1

)|2

+ h(1 +1

γh)E

∫ tk+1

tk

|f(tk, XNtk, Y N

tk+1, ZN

tk) − f(s,XN

s , YN,1s , ZN,1

s )|2ds

≤(1 + γh)E|Etk(YNtk+1

− Y N,1tk+1

)|2 + C(1 +1

γh)h2 sup

tk≤s≤tk+1

(E|XNtk−Xs|2 + |tk − s|)

+ C(1 +1

γh)h2 sup

tk≤s≤tk+1

E|Y Ntk+1

− Y N,1s |2 + Ch(1 +

1

γh)E

∫ tk+1

tk

|ZNtk− ZN,1

s |2ds

≤(1 + Ch)E(|Y Ntk+1

− Y N,1tk+1

|2) + Ch2 + ChE

∫ tk+1

tk

|ZN,1s |2ds+ CE

∫ tk+1

tk

|ZN,1s − Z

N

tk|2ds

@ EN¡KLK6"O 533QECh

"K6suptk≤s≤tk+1

E|Y N,1tk+1

− Y N,1s |2 J¤

EN3!E|Y N

tk+1− Y N,1

s |2 ≤ 2E|Y Ntk+1

− Y N,1tk+1

|2 + 2E|Y N,1tk+1

− Y N,1s |2 ±¶E"!T DE£DEQEJ ¬´\ dMN LE5! E£ KLEJ )!

E∫ tk+1

tk|ZN,1

s |2ds3DDQEKLK6¡!:µ"¬E ¶! D$ ±YD EFE6K6C M

Page 141: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸

L2 !"FE)D! K6QE E|Y N,1

tk+1|2 3QE)6EDEDE:!¡/KL! !

1 MV /E MNPP K6KL!µ"¬E T!D· ¥ E¬DN - ±YK6KLO :MNY3!

suptk≤t≤tk+1E|ZN,1

t |2 ≤ C(1 + |X0|2) OP =NQQEJKLN

E|Y Ntk

− Y N,1tk

|2 ≤ (1 + Ch)E(|Y Ntk+1

− Y N,1tk+1

|2) + Ch2 + CE

∫ tk+1

tk

|ZN,1s − Z

N

tk|2ds.

±Y¡K K6¦V!DED MV!QE3E:QE "´K6·3! EJ ¬FEH¡N−1∑

k=0

E

∫ tk+1

tk

|ZN,1t − ZN

tk|2dt.

¤OP FE 7!7DND¶ a !cY EY K6E∫ tk+1

tk|ZN,1

s − ZN

tk|2ds E EV !¬DD!

(Y N,1, ZN,1)

(Y, Z) ª c!)V c! !N!¡FEH E ! EJD

g(·, ·)

(+&'! #(C1,2

¨'^E Dv# NV> v`7f-, Z@ .S%A ACA

Lg !#, +

C1,2 + *! # $ '$ #$ ª @V%! ^E 56NV! I v LK6L! E MV5 EJ$ Q°7=DK:QE¡±YK6K6"O ·3" K6O =NMN

E sup0≤t≤T

|Y N,1t − Yt|2 +

∫ T

0

|Zs − ZN,1s |2ds ≤ C(1 + |X0|4)

√h.

­P¡Q=3KLE ∑N−1k=0 E

∫ tk+1

tk|ZN,1

s − ZN

tk|2ds QE

CE

∫ T

0

|ZN,1t − Zt|2dt+ C

N−1∑

k=0

E

∫ tk+1

tk

|Zt −1

hEtk

∫ tk+1

tk

Zsds|2dt

≤C(1 + |X0|4)√h

DKHQE¡5KLK6:O L·: K66O ¡5K6K66g E¡ K6¡O :"D!@KL OP !QE"DEg

3!¡DEC1,2 E£D!3! 3 $</

max0≤k≤N

E(|Y Ntk

− Ytk |2) +N−1∑

k=0

E

∫ tk+1

tk

|Zt − ZNtk|2dt ≤ C(1 + |X0|4)

√h.

Page 142: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP±\µ¨"³´ «I%°¥d¨"©P³±\° ¸ °7­ ¸ ³3³,'SVegUh ¶S%ACACA

$ ! +- +-$ + *! , $ ! $"+ , + ! $ #

& #$ ! "&- + √hh

14 # , + #$ ! , +'!#,

g , %$ +, #$ & #$ # +-,

¯J¡! E MND·PE K6"E=D£EJ3! EKL D£E =# ED" QE¡!5DEC1,2 DKLK6:QEJ¡KL 5!QE 5DE¡! QEC

(K − ex)+

¶¨"Y¡KL¤ MN¬¬E NV¬ ED g

¬ EK6D DD ! 3!QE I v # + ,2 #Y-+* "\% #" +*#+* *#& B *; #:<¨'^E !QED·¡QE # NV JE> v`7f-, Z@ .S%A ACA

O $ g !#, +

C1 #$ ' '$ #$ !' #$

x

g(t,Xt)g(t,XN

t )+ $ , #, $

g(t,Xt) =g(0, X0) +

∫ t

0

Usds+

∫ t

0

VsdWs + At,

g(t,XNt ) =g(0, XN

0 ) +

∫ t

0

UNs ds+

∫ t

0

V Ns dWs + ANt ,

A AN $ ! $ %$ !# + $ %$ "&-)+, #, , dAt

dANt

$#%$'& ', # +- +# dt + ! , ! $ - $ %$ "&- +%, +-$ supt E(|Ut|p) +

supt E(|UNt |p) + supt E(|Vt|p) + supt E(|V N

t |p) <∞ ' p ≥ 2

³KLEMN'MV¬QEK6 ' %D£E'! /EKL D£E !K6 1 # NVJ¤¡´´´ : Q ¡D£E

g(t, x) = ψ(t, exp(x))E£D

ψ(t, ·) D <OP / aK:! ´º 6 EN ¬ K6%E QE !QE ² ¸ G

g(t,Xt) = ψ(t, exp(Xt)) =g(0, X0) +

∫ t

0

Usds+

∫ t

0

VsdWs + At

g(t,XNt ) = ψ(t, exp(XN

t )) =g(0, XN0 ) +

∫ t

0

UNs ds+

∫ t

0

V Ns dWs + ANt

ECDtk ≤ t < tk+1

Ut =∂tψ(t, exp(Xt)) + ∂xψ−(t, exp(Xt)) exp(Xt)b(t,Xt) +

1

2σ2(t,Xt),

UNt =∂tψ(t, exp(XN

t )) + ∂xψ−(t, exp(XN

t )) exp(XNt )b(tk, XN

tk) +

1

2σ2(tk, X

Ntk

),

Vt =∂xψ−(t, exp(Xt)) exp(Xt)σ(t,Xt),

V Nt =∂xψ

−(t, exp(XNt )) exp(XN

t )σ(tk, XNtk

).

Page 143: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸ª K6K6

∂xψ− ' %!E =D£EL!'EK6 D£E ! KL

1d LD! ! $E !¡# NV´´´ : Q $

±YMN ® EKL ¡3P! P!E I v #<% EJT MNE5D! g

!DEC1,2 VNN KLN(!6:! "G

­EH6K6K66O b±¶E% $E 5!g

H 6)K6)DN ) VDdDEK

KN,0 N¨$D vK6dDMN 5 KL¬!QEc# NV´´´ <± MN /!5 EJ MNE:KLEMN¡O 5MN¶!

0 ≤ dKt ≤ f(t,Xt, g(t,Xt), Vt) + Ut−dt = ktdt,^´´´

0 ≤ dKN,0t ≤ f(t,XN

t , g(t,XNt ), V N

t ) + UNt −dt = kN,0t dt,

P%P KK6DN PPVD"DEPMN)!QEPD£EP g

!¡DEC1,2 <°c=QE D $

supt,N E|kN,0t |p <∞ ·supt E|kt|p <∞

p ≥ 2

­E3 K6¡O Q­E3¡DE$Q=EE:KLN3 " EJ

E sup0≤t≤T

|Y N,1t − Y N,0

t |2 +

∫ T

0

|ZN,1t − ZN,0

t |2dt ≤ C√h

"¡!V EEJE- K6"O a ^O EH! Q -!(Y N,0, ZN,0)

Yª D ¡ ^E QEP EC)VFE6V :! HQEP) K6K6O 7CDK6KLNK6D¡ EJ¡G ±¶E/6 MVLK6 EJ !/D :! I v # Y¯¡E-!¡!¡KK6"!QE@ N $E4 ! =DE5D QE) ¨:

∆Yt = Y N,0t −Y N,1

t

∆Zt = ZN,0

t −ZN,1t

HKEMN¬MN∆Yt

EPE!VQEK6MN JE¡]tk, tk+1[

−d∆Yt = [f(t,XNt , Y

N,0t , ZN,0

t ) − f(t,XNt , Y

N,1t , ZN,1

t )]dt− (ZN,0t − ZN,1

t )dWt + dKN,0t .

OP %EJ MNQEE aKHP! ´º 6= Nd|∆Yt|2 = − 2∆Yt[f(t,XN

t , YN,0t , ZN,0

t ) − f(t,XNt , Y

N,1t , ZN,1

t )]dt+ 2∆Yt∆ZtdWt

− 2∆YtdKN,0t + |∆Zt|2dt.

ª (!5 NC > 0

YMN

2|∆Ys||f(s,XNs , Y

N,0s , ZN,0

s ) − f(s,XNs , Y

N,1s , ZN,1

s )| ≤ C|∆Ys|2 +1

2|∆Zs|2

^´´´

DMV dH EJN\MNf

c±Y D d $E 7! d*^O O#®NP E MNQEJNP aKH! ´º !Λs|∆Ys|2

[t, tk+1]

ECDΛs = exp(Cs)

Page 144: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ª¬«O¥c´ ³3° @OP±\µ¨"³´ «I%°¥d¨"©P³±\° ¸ °7­ ¸ ³3³^

C3FEHD EJN!¡# MNQE/ D ! ¡G

Λt|∆Yt|2 +

∫ tk+1

t

Λs|∆Zs|2ds

=Λtk+1|∆Ytk+1

|2 + 2

∫ tk+1

t

Λs∆Ys[f(s,XNs , Y

N,0s , ZN,0

s ) − f(s,XNs , Y

N,1s , ZN,1

s )]ds

− 2

∫ tk+1

t

Λs∆Ys∆ZsdWs + 2

∫ tk+1

t

Λs∆YsdKN,0s − C

∫ tk+1

t

Λs|∆Ys|2ds @ E a´´´ G

Λt|∆Yt|2 +1

2

∫ tk+1

t

Λs|∆Zs|2ds

≤Λtk+1|∆Ytk+1

|2 − 2

∫ tk+1

t

Λs∆Ys∆ZsdWs + 2

∫ tk+1

t

Λs∆YsdKN,0s .

´ \P !LDN 3!3KL!®KLKH!)! %VH!)DD QE=KLK6!µ"¬E !D Q¥bDFE@E-G∆YsdK

N,0s =[g(s,XN

s ) − Y N,1s ]dKN,0

s

=[g(s,XNs ) − EsY N,1

tk+1+

∫ tk+1

s

f(r,XNr , Y

N,1r , ZN,1

r )dr]dKN,0s

≤Es[g(s,XNs ) − g(tk+1, X

Ntk+1

) −∫ tk+1

s

f(r,XNr , Y

N,1r , ZN,1

r )dr]dKN,0s .

ª" % KLK6N=FE! K6EJ D QEJ 3°7 ·%=DN /FE! Dg(s,XN

s ) − g(tk+1, XNtk+1

) =E@E MVN/ aKH ®! ´º oD MN¡ /K6 EJ)D

Ch!QEb# MNQED ¤º!$ J­EJbdD£Ebb!dK K6TD·¡:KE EJ ® N EJN YO#T E*^´´´ ¡ V5´´´ E MV " K6KLPO H!= MV v 3¬JEE # NV´´´ Q/¡G

E

∫ tk

t

Λr∆YrdKN,0r

≤CE

∫ tk

t

Λr|r − tk| + |XNr −XN

tk| +∫ tk

r

|f(s,XNs , Y

N,1s , ZN,1

s )|dskN,0r dr

≤Ch 32 .

¨=NQQEJKLN¡GE[Λt|∆Yt|2 +

1

2

∫ tk+1

t

Λs|∆Zs|2ds] ≤Λtk+1[E|∆Ytk+1

|2 + Ch32 ]

≤Λt(1 + Ch)[E|∆Ytk+1|2 + Ch

32 ].

Page 145: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´´´ <­P° ¸ ° »P°d¯ ¸ ´ ¨"¯ ¸ª K6KL

Λt

D EN$Q VN¡G

E[|∆Yt|2 +1

2

∫ tk+1

t

|∆Zs|2ds] ≤ (1 + Ch)E|∆Ytk+1|2 + Ch

32

3EJ D£EJ '!@±¶K6KL"!)µ"¬E ! D3K6¡E= EJ$ ¤¨6 EJ QcMV7dEJ7 EJ7!QEJ I v Q NcQEJ7DEED·

C1,2 !g OP#=EE:"KKL"D

max0≤k≤N

E(|Y Ntk

− Ytk |2) +N−1∑

k=0

E

∫ tk+1

tk

|Zt − ZNtk|2dt ≤ C

√h.

#

¸ c K66 E · E"!QEJDLD QEJ Y !D6 ¡NV´´´ Q´´´ 5·3 ´´´ ´´´

max0≤k≤N

E|yN,R,Mk (XNtk

) − Ytk |2 +N−1∑

k=0

E

∫ tk+1

tk

|Zt − zN,R,Mk (XNtk

)|2dt

≤C√h+ C

N−1∑

k=0

E[1g(tk,XN

tk)6=gR(tk,X

Ntk

)|Y Ntk|2 + Cy(R)2

]+ CE(|g(tN , XN

tN) − gR(tN , X

NtN

)|2)

+ ChE

N−1∑

k=0

|f(tk, XNtk, Y N

tk+1, ZN

tk) − fR(tk, X

Ntk, Y N

tk+1, ZN

tk)|2

+ CCy(R)2 + ‖fR‖2

∞M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + Chβ−1 + CN−1∑

k=0

infα

E(|yN,Rk (PN

tk) − α.p0,k(P

Ntk

)|2)

+ CN−1∑

k=0

q∑

l=1

infα

E(|α.pl,k(PN

tk) −

√hzN,Rl,k (PN

tk)|2)

+ C(Cy(R)2 + ‖fR‖2

∞) maxlKl,k

Mlog(M)

+CCy(R)2

hexp(− Mhβ+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

Page 146: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

sl-t=pv| wkr l@p q|

Page 147: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 148: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

kst=l%pr | j.syl't@pv|

-kt@st@ wkr l@p q|

Page 149: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 150: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

syjkj | q| ~ksj.pt=l@| jkl ~ q|w1t@

­E3D·¡ <! 3EVD FEH -!¡# °7­ ¸ ³xG

Yt = Φ(X) +

∫ T

t

f(s,Xs, Ys, Zs)ds−∫ T

t

ZsdWs

^´ /

W'KLK6!%! KL

qX

'FE o!%# MNQEJ! VD QEJ MN@! KL d

GXi,t = Xi,0 +

∫ t

0

bi(s,Xs)ds+

∫ t

0

σi(s,Xs)dWs, 1 ≤ i ≤ d.

Φ *aD· :!55FE'ED:!6FE-!T

SYDDd"¡DK6! bQE¤º!! \D QE ¬QE ´ J¥b ! EEJc!:DEEJD QEJV¤!!N$:EVD ¬FED! KL QE

Φ(X)QE

ΦN(PNtN

)J

ΦN 7 aD! ·KL 3(PN

tk)0≤k≤N

¬PD QEP!"I%E L!P!K6d′ ≥ d

!N d

·¤K67DKLENd ! QEJdDd!(XN

tk)0≤k≤N

N°7L! EJdK6V'E N KLN ! EJJEFE d! EJ¬ !PKE V E!VQEK6MNK6 D !¡FEHD! 'K6E ¨@E5 ·! ¡!EQE ´3´´ !@EJ K6¬ ·EJ!3EVD FEH ¤ ®!^´ / ¶ª¡! EJ K6PN¡E ¡"!¡ K6!:K6 !DE MN ® EJ K63L! )!D · EJ =@K6! NJEJ

[0, T ]a

N +1!QEJ !! D ELN J

(tk = kh)0≤k≤N

h = TN

°c' ·V! EVD "FE'(Ytk , Ztk)

!-D QEMN5 E¡!5! D EJ tk

E"!¡D· P!PQE! aD· T±¶ELQEJ) PEVD Ytk

J p0,k(P

Ntk

) = p0,k5q

QE5 H5EVD 5D QED6!q

DK6EJN5!Ztk

N5 (pl,k(P

Ntk

) = pl,k)1≤l≤q <±Y¡DE! QEv!E:E

pl,k J

Kl,k

Page 151: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ 7³° ¸ ©±\Oc ¸ ¯©I%°7³´ "©P° ¸³3E P) MV:K6P!HKL !PDE "N"E ¡!"KHEJ !I@ ¤ ª¬E%!FE@D QE !/I@E C

(PNtk

)0≤k≤N ±Y'KHL!'KHFE !-I@N·¤ª¬EJ + P KHFE N

(PN,mtk

)0≤k≤N,1≤m≤M v±Y"E!3KEJ D¬! $@!¡D K6!K6 !D£EJ EVD "!

pl,kE

M KHFE ¬N3 Kl,k(ω)

O\EN !d b! HE K6\ N !QEbbQE b´v ´´ b! DQ"D -! aD·!¡QE

Page 152: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

ksyj.pt=l=|

qmpQn q| 7mwk~t@pvmwk q| .sx'|

­E5FE-E:VK6 MN ¶¡!T N¡ !*aD "!6QE \¥ EDQ!DE dEC7$T E"!D 5!Dd! c Vd!aD· 5!QEJFEHKK6"QE

¥ $ pl,k

l

k ¥!!"$#% &'()*+'-,.(!/0D ⊂ Rd′ *'1(2

P0

* 3 D =

∏d′

i=1 ]P0,i −R,P0,i +R]4125 6().()7+ ((5,8 '#9:(

δ

; D = ∪i1,··· ,id′Di1,··· ,id′

Di1,··· ,id′ =]P0,1−R+ i1δ, P0,1−R+(i1 +1)δ]×· · ·×]P0,d′ −

R + id′δ, P0,d′ −R + (id′ + 1)δ] ¯<75

pl,k(·)'09" .= &'(0'>('?@' !A'(

B#CD,N@ '#FEpl,k(·) =

(1Di1,··· ,i

d′(·))

i1,··· ,id′ ªG(()H#C%&'((I(HJ2K5

­L< $@M'G·N()N(GM'-,41L,8 '#OG @P@'>Q ((P V¥RP80 NP M <I#II @0P@'>Q'(()

1, x1, · · · , xd′

p0,k

(1

pl,kl ≥ 1

;

p0,k(x) =(1Di1,··· ,i

d′(x), x11Di1,··· ,i

d′(x), · · · , xd′1Di1,··· ,i

d′(x))

i1,··· ,id′,

pl,k(x) =(1Di1,··· ,i

d′(x))

i1,··· ,id′, 1 ≤ l ≤ q.

ªGQ'S,% ()'L &'T()OQ#Q(H()J%K ?> H1

0L N()0

C0 80UC"B#HC @@'! p0,k

pl,k

1 ≤ l ≤ q

Page 153: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

´ 7³° ¸ ©± ; ¸ ¯©%°7³´ "©P° ¸

´ 'G!'5'%%#2 &'T()7'()'?(!"$ #%

0 ≤ l ≤ q vªGT(() &$L # "&'T()<'> ()'<N(OB 6().()9(

¹ µ¡±¶ N(P '1()M1(PPB( 1(MPN N¥ ·¤

'01($1 MM P 0 (M<#I20

&'( '>('$NB

20()'( 1(-I

PN ()(PN,M+i)1≤i≤20

0N(

(PN,m)1≤m≤M V¥ $NP() !H3()N(

tk (PN,M+itk

)1≤i≤20

P () ¤().()9 ¹

(Ck,i)1≤i≤20

Ck,i = x : |x − PN,M+i

tk| < infj 6=i |x − PN,M+j

tk| ° )

R pl,k(·) =

(1Ck,i

(·))

i

ªG(()G#G &'T() (R() V¥R(()- ­L* G41 <#J%K ' 1(01( '-,41'M J¤¹ 8 0 @'OP T()( $ ¥ M 4@M J%K5 (F'HH'-,410' # 8 " '6().()0

1, x1, · · · , xd′ p0,k

(1

pl,kl ≥ 1

*'S,41Ck,i

;

p0,k(x) =(1Ck,i

(x), x11Ck,i(x), · · · , xd′1Ck,i

(x))

i,

pl,k(x) =(1Ck,i

(x))

i, 1 ≤ l ≤ q.

¯<O()O'(() # ()' > O(1, 0)

?"?0 '() 41 J2K5

#

¯<*'p0,k

'0<"F# @Gd′

"#MT( )&!dy

(pl,k

l ≥ 1

*'0 "B#HC 8 0C$C)&!dz

Page 154: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

"! # $ %#

&<O£L 0 ()(')'() 2+* (),H41().-,0/.D6(-N(tk0B" ().

#.D'01(7#10!328"546&I!£701( ! 7'-, () 'T(( 6()5? (041I0N()1(041#.D6(7 #=#.3 410=8799*.(,0*IB" ().!:.3'N()#10<328<;T(< ##01(HC#I#.D<41 1(L4@1(I=B#C>=&'(C#?;

M(I

h;=87 (A@8 C41H'(-#(), B')'C4ED3;F'G'C4IHJ;F'G'C4LK0T(M'G'C4ON34&<C<'' L/.3+*.(,F N()QO? ()-'G'<C' %OB" ().#.D'01(I#P0<328<4Q '

d = q = 1T(F '

S"!J@@41Q @ 8RM"'CSUTWV8'-,

XGY RZV\[?H?]_^ E

dStSt

= µdt+ σdWt.

` F'(Q()</.Iacb>VJd/6((ST −K)+

;@'I#.37'> #4FeGI f;5 E

f(t, x, y, z) = −θz − ry + (y − z

σ)−(R− r)

C'θ = µ−r

σ

; OH4@f4` ) T()H 3@8041C

ST(IC/.D ()!1(CE

µ σ r R T S0 Kg 4 g N g 4OD g 4 g K g 4 gUh g 4EN i gfg i gUg

e6j((kjml)n.(koFprqkn+sUtvu5sUw)wGj<lGxsUt\yzpfou|pfl6sU~9j3Uj<tJy\j<o\wZyJjm~/.LsUxJq)n9sftj<8x\w)o\tqkjy\j~/.Ipfw)*Uj5tqpfovqkpfo32y#.Ln9tqk<wG75q

Rj5q(x\wG75q)jvpfoqCpfo32

rpCUj5u

r < RXGY R6j5wkNr]_^4e6j5qGqkjn+xj5wG=j5u5qkn+sUty\j8pfw)uC\xFpfw

Page 155: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

wkp?x\xsUwGqcp?o32 @3xs?qk5l)j<l n+y\|pf~+j<l y\j Y RV\[fH?]\u5w)<jZo\t\jZt\sUt TA~9n+t\<pfw)n q)6y\pft\l ~9j6y\w)n Uj<wf;o\n\l)n+t\sUt?p?o\y\wkp?n+qBlGn98x\~9j58j<tq

f(t, x, y, z) = −θz − ry.&MsUo\lMpCfsUt\lBu JsUn9lGnu5j5qMj23j5x\~+j u<pfwB~+j>y\wGn Uj<w

fj5lGq6t\sUt3T ~9n+t\|p?n+w)jBpfn+l tJsUo\l6xsfoUsUt\l6t\<pft\8sUn+t\l u<sUtJtFpfn+q)w)jM~_p"?pf~+j<o\w

Y0

4 a tj j5q<;3~+jMy\w)n Uj<wfl)j wG<5u<w)n q

f(t, x, y, z) = − µ−R

σz −Ry + (y − z

σ)+(R− r)

≥− µ−R

σz −Ry.

e6j>yJj<w)tJn9j5w6qkj5w)8jmqCpftqZ~9j y\w)n Uj<wu5sUw)wGj<lGxsUt\yFpftqM,-o\t\jmyJ@ tFpf8no\jmyJj>xsfwGqkj=j<o\n+~9~+j u5~_pflGl)no\jpCfj<uo\t o\tJno\jqCpfo 2 y#.Ln9tqk<wG75q

R;sUt l)pfn+q(j<t oJqkn+~9n+l)pftq(~+j8qkJ<sUw<8jvy\jvu<sU8xFp?wkpfn+l)sUt xsUo\w

~9j5lvacb>VJd XGY a>[r]_^o\jY0

l)j5wkp lGo\x<wGn9j5o\wpfoxJw)n 2 u<pf~9u5o\~9zxFpfw~9p=sUwG o\~9jzy\j R6~9pfuCSTV3uC\sU~+j<l6pCUj5u qkpfo32y:.In+tqk5w)7q

R4pfn9l sUtlkpfn q <*p?~9j5j5tq y\pft\l ~9jMu<pfl6yJj~9p wG<x\~+n9u<p?qkn+sUtyFpft\l ~+j

8s y<~+jMR6~_p?u SUTWV u JsU~9j5lo\jxsfo\w u<sUo3wGn9w ~ .IsfxJqkn+sUt#;U~+jUj5t\y\j5o\w6y\j w)p(q)sUoGsUoJw)l j<8x\wGo\tq)j<w , ~_p pft oJjf;3u?.Lj<lGqWT ,rT y\n9wGj!o\jf; l)n sUtvtJsfqkj

(Y BSt , ZBS

t )~+j<l l)sf~9oJq)n9sUtJl yJj ~/.Ia bmVJd pfl)lGs u<n+<jpfoy\w)n Uj<w

−µ−Rσz − Ry

; pf~+sUw)lY BSt ≤ ZBS

t

σ

; ZBSt

σ

w)j<xJw)<lGj<tqCpftq~9j>sUtqCpftqBn9tUj<lWqkn j5t p?u5q)n9sUt#4 n9t\lGn/;\sUts q)n9j5tq n+j<t ~ .I5*pf~+n q) y\j<l yJw)n Uj5w)l

fj5q −µ−R

σz − Ry

~+j ~9sUtJ*zy\j~_pl)sf~9oJq)n9sUt(Y BS

t , ZBSt )

jqy\sUtJu ~ .I5*pf~+n q)

Y0 = Y BS0

4Y BS

0

lGj u<pf~9u5o\~9jmj23x\~9n+u<n qkj<8j5tqZj5qrpfoJq7.156

4Q t qkj<lWqkj n+u<nB~_p pfl)j#"%$4 Q tt\s?qkj

K~9jt\sU wGjy\ju<j5~9~9oJ~9j<ly\j#csfw)sUt\sfnMo3qkn9~+n+l)5j<l54& sUo\w

u<sfxFp?w)j5w ~9j5l p?~9*UsUwGn qkJ8j<l p£Uj<usUo lkpftJl lGn9 o\~_prqkn sUt(y#.Ln9t\u5w)5j5tq)l w)s10Bt\n9j5t\l pfo323n+~9n9pfn w)j5l<;rt\sfo\ly\5u<n9yJsUt\lBy\j>= p?n9w)jPrpfw)n+j<wl)n9 o\~ qCp?t\<8j5tq

N;M

j5qK4 a txFp?wGqkn+u<o\~+n+j<w5;JtJsUo\lZ= pfn+l)sUtJl¬?pfwGn9j5w

Ny\j q)j<~9~+j-l)sfwGqkj' oJjN = [N0(

√2)j−1]

sfj = 1 . . .

yJ<l)n+*Ut\j ~+j t\sf w)j y\jHrpf~9j5o\w)lmy\jN

o\j~/.LsUt qkj5lGqkj X

[·] y\<lGn9*Ut\j n9u<n ~_pxFpfwWqkn+j j5tqkn(<wGj1^4a t\l)oJn+qkj?;t\sUo\l>= p?n9l)sft\l?pfwGn9j5wK

j5qM

y\j q)j<~+~9jl)sfwGqkjo\j

K = [K0(√

2)(j−1)αK ]j5qM = [M0(

√2)(j−1)αM ]

p£Uj<uN0 = 2 = M0 = K0

4 sUo\wu \po\j=?p?~9j<oJwy\j

(αK , αM)t\sUoJl8w)j<xsUwWqksUtJl8~9jzxJw)n 2 8sP@Uj5t y\sUtJt\xFpfw~/.Ipf~+*UsUwGn q)\8j

X u<pf~9u5o\~+Bj<t~_p?t)|pftq6N g =sUn+lc~ .Epf~+*Usfw)n qk\j1^;f~/.L<u<pfwGqWT qA@ x jBy\jMu<jMx\w)n 2 ;~_p>u<sUoUj<wGq)o\w)j 8sP@Uj5t\t\jMj<tt0jqM~/.L<u<pfwGqWT qA@ x j>y\j ~_p u<sfoUj<wWqko\w)j?4*csUn9u5n#~+j<lBwG<lGo\~+qkp?qklBs qkj5t o\l

j

Prix

105

5

6

7

4.5

5.5

6.5

7.5

Prix avec browniens auxiliairesPrix sans browniens auxiliairesPrix Black-Scholes

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

+-,/.

Page 156: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Bd a"HJ4 ma e Q V m&VRd Q &M'Wac&V%!m' `' 'WdMa V

j

Couve

rture

105

0.2

0.3

0.4

0.5

0.6

Couverture avec browniens supplémentairesCouverture sans browniens supplémentairesCouverture Black-Scholes

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105

1

2

3

4

5

6

7Avec browniens auxiliairesSans browniens auxiliaires

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105

1

0.5

Avec browniens auxiliairesSans browniens auxiliaires

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 ! "$#&%$'()* +*,-.#!( !/%-#&0213#!*"$4*(5*%$6"$+7#!*

989#!#!#&"$*:$*(6!*% $;*()<=".%-*">?@A()B()?(5#C".ADE/1F*.% 9"$#G<H">I%I89()<!?I%4 !*(5*%JK"$#L<H"::%-".:I89()<>#L$#&M%ONQPR#!$#L<":**#!$#2STU/:%$%$5@$#%-/SU<"$4*$*%-V%$W%-OJX1Y/! ".#&%-'()Z$$#!( !/%$#213#!"$4*(5*%$6"$+7#!*989#!#!#&"$*N

[\\

Page 157: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

iPN h

Page 158: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

# %

! "#!$&%(')+*-,.)0/1/1)324/5'6,7"98:8/;$=<> "#!$?%@8A$&2/B*"DC"&)EGF

& sfo\lMpIHUsUt\lBs qkj<toyFp?t\lM~9p x\pfwGq)n9j>'G' ~9j>w)5l)o\~ qCp?qBlGo\nJHrpftq<4

max0≤k≤N

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

+ hE

N−1∑

k=0

1

M

M∑

m=1

|zN,R,Mk (PN,mtk

) − zN,Rk (PN,mtk

)|2

≤CCy(R)2 + ‖fR‖2∞

M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + CN−1∑

k=0

infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2

+ CN−1∑

k=0

q∑

l=1

infα

E|α.pl,k(PNtk

) −√hzN,Rl,k (PN

tk)|2 + Chβ−1

+CCy(R)2

hexp(− Mhβ+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

X ' 4LK\4 iP^dMpfx\xj<~+sUt\l X HUsfn9wMxFpfwGq)n9j ')'W^ oJj (

yN,R(PN), zN,R(PN)) j<lGqm~_p8lGsU~9oJq)n9sUty:.Io\tJj y\n+l)u<wG5q)n9lkprqkn9sft

j<tqkj5x\lcyJj X '-4+if4+iP^ j5q y#.Lo\t\jM~+s u|pf~+n+lkprqkn+sft y\qkj<wGn+t\<jZxFpfwcoJtKHUj5u5qkj5o\wR4 e6jq)q)jM~9s u<pf~9n lkp?qkn sUt

t\sUoJlZxj<wGjqZy\jmsUtqkwGj<w o\jyN,R(PN)

j5lGq sUwGt\mxFpfwoJt\j>u<sft\lGqkpftq)jCy(R)

o\n#yJ<xj<tJyy\j~_p ~+s u|pf~+n+lkprqkn+sft#4d pfxJxj<~9sft\l j<tMLFt oJj~/.Ipfx\xJw)s123n+p?qkn sUtw)5l)o\~ qCpftq6yJj ~/.Ipf~9*fsUwGn q)\8jByJj ~_p xFpfwGq)n9j'G'6j5lGqMt\sfqk5j (

yN,R,M (PN), zN,R,M (PN)) 4

& sfo\lMpf~9~+sUt\lyJ<y\o\n+w)j y\owG<l)oJ~+qCprqMu5n T y\j<lGl)o\lM~9p x\wGsUxsUlGn+qkn+sUtlGo\nJH?p?tq)j

Page 159: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

" !#"%$&'()$+*,$"-/. $0$&1&-,23#-,"546$&78$&9;:<=7>-,"δ

-<M

?-<@"BA%'-δ = hγ

4'C7γ > 1

2

-<M = h−(β+1+γd′) $&1&-,2

β > 1 DCEC&*%F-<'HG

maxk

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2 + hE

N−1∑

k=0

1

M

M∑

m=1

|zN,R,Mk (PN,mtk

) − zN,Rk (PN,mtk

)|2

≤o(1) + C(R)N−1∑

k=0

P(PNtk/∈ D)

CIo(1)

?-<J;1&-<7>"0

A%6$&Jh

K-<JL1&-<7>"0 M

N(OQPRTSUPWV a tpfx\x\~+noFpftq X ' 4LK\4 iP^pIHUj<u o\txFpfwkp? qkw)jβ′ qkj5~ o\j 1 < β ′ < β

sUtzs q)n9j<tq

max0≤k≤N

E1

M

M∑

m=1

|yN,Rk (PN,mtk

) − yN,R,Mk (PN,mtk

)|2

+ hE

N−1∑

k=0

1

M

M∑

m=1

|zN,R,Mk (PN,mtk

) − zN,Rk (PN,mtk

)|2

≤CCy(R)2 + ‖fR‖2∞

M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω)) + C

N−1∑

k=0

infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2

+ C

N−1∑

k=0

q∑

l=1

infα

E|α.pl,k(PNtk

) −√hzN,Rl,k (PN

tk)|2 + Chβ

′−1

+CCy(R)2

hexp(− Mhβ

′+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ′+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

sUoJq y#.Ip sUwGy#; t\sfo\l l)pIHUsUtJl' oJjyN,Rk (·) jq (

√hzN,Rl,k (·))1≤l≤q

l)sUtq y\j<l-=sft\u5q)n9sUt\l-`n9xJl)uC\n+qYXo\t\n =sUw)8<8j5tqZj5t

kjqN4* n9t\lGn/;JxsUo\wB~9p pfl)j. $;\t\sUoJlMs qkj<tJsUt\l

E(Rp0,k

(Y N,Rtk

)2)

≤E(|Y N,Rtk

|21Dc(PNtk

))

+∑

i1,··· ,id′E(1Di1,··· ,i

d′(PN

tk)|yN,Rk (PN

tk) − yN,Rk (xi1,··· ,id′ )|2

)

≤ Cδ2 + Cy(R)2P(PNtk/∈ D),

sUxi1,··· ,id′

j<lGqo\t x sfn9tqpfw n+qkw)pfn+w)jyJjDi1,··· ,id′

j5qsU tJsUo\lpIHUsUtJloJqkn+~9n l)~_pxJw)sUx\wGn9qkzy\j`n9x\lGuC\n+qYX yJj

yN,Rk

lGo\wD4 a txsUlkp?tq , x\w)5l)j<tq

η = δh−12 → 0

xFpfwZ@ xsfqk<l)j?;sUts qkn+j<tq

infα

E|α.p0,k(PNtk

) − yN,Rk (PNtk

)|2 ≤ Chη2 + Cy(R)2P(PNtk/∈ D).

iPN&Z

Page 160: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

bj5l u<pf~+u<o\~+lZl)n98n+~9p?n+wGj<ll)sUtq=H?pf~9p ~+j<lZxsUoJwE(Rpl,k

(√hZN,R

l,tk)2) jqMsUtj5ty\5y\o\n q

maxk

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2 + hE

N−1∑

k=0

1

M

M∑

m=1

|zN,R,Mk (PN,mtk

) − zN,Rk (PN,mtk

)|2

≤C(hβ′−1 + η2) + CCy(R)2

N−1∑

k=0

P(PNtk/∈ D) + C

Cy(R)2 + ‖fR‖2∞

M

N−1∑

k=0

max0≤l≤q

E(Kl,k(ω))

+CCy(R)2

hexp(− Mhβ

′+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ′+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

Kl,k(ω)j5lGq p?~9sUwGl pGsUwG6xFp?w ~9j6t\sU w)jy:.I@ x j5w)u5o j<l y#.Ipfw)7qkj

δu<sftqkj5t oyFp?t\l ~+jyJsUvp?n9t\j

D4

% Q 'u|pfw l)j<oJ~9l u<sU8xJq)j<tq6~9j5l @ x j5w)u5o j<lyFpft\l ~9j5l o\j<~+l l)sUtq6x\w)5l)j5tqkj5ly\j5l-lGn9 o\~9p?q)n+sft\l

(PN,mtk

)1≤m≤M;u<sU88jsUt ~/.Ipzy\ ),n+t\y\n( oJ8yFp?t\l(~9pzxFpfwGq)n9j'G'C4 n+t\l)n ;sUt p

Kl,k(ω) ≤ (2Rδ

)d′ l)sUn q

Kl,k(ω) ≤ C(R)δ−d′ H3o ~9j(uC\sUn 2y\j

Rj5qδj :j5u5q)o\<l>x\wG<u5<y\j58j<tq|4

'A~ H n9j5tq

maxk

E1

M

M∑

m=1

|yN,R,Mk (PN,mtk

) − yN,Rk (PN,mtk

)|2 + hE

N−1∑

k=0

1

M

M∑

m=1

|zN,R,Mk (PN,mtk

) − zN,Rk (PN,mtk

)|2

≤C(hβ′−1 + η2) + CCy(R)2

N−1∑

k=0

P(PNtk/∈ D) +

C(R)δ−d′

Mh

+CCy(R)2

hexp(− Mhβ

′+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ′+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

Q toJqkn+~9n+lGj ,x\wG<l)j5tq ~9j>u<sUtqkwU~9j>y\ot\sU w)j>y\j u5sUoMHUj5wGq)o\w)j N2

j<t=sUtJu5qkn+sUtzy\jδj5qhyJsUt\t\

xFpfw X 'G' 4IHJ4LK^ X HUsfn9w xFpfwWqkn9j('G'W^4#e6sU8j(sUtpvuC\sUn+l)nM ≈ h−(β+1+γd′) pIHUj<u β > β ′ sfts qkn+j<tqLFtFp?~9j<8j5tq oJj(~+j<lBy\j5o32zqkj<wGj5lMn+xJ~9n oFpftq

MyFpft\lM~ .In+t\ o\p?qkn+sUtzu5n T y\j<lGl)o\lBqkj5t\y\j<tq HUj5w)l

04

¤

& sfo\l y\<lGn9w)sft\l?HU<wGnJLFj5w ~_pH?pf~+n9y\n q)y\ju<jq)q)jx\wGsUxsUl)n qkn+sUt#;u?.Ij5lGqWT ,rT y\n9wGj HUsUn9w(l)n6j<t = pfn+lkpftq?HrprTw)n+j<w

N, δj5qM

y\j qkj5~9~+j l)sUwWqkj oJj-~+j pGsfwkpftqMyJj(~_px\wGsUxsUlGn+qkn+sUtzq)j<t\yJj&HUj5w)l0; ~/.Ipf~+*Usfw)n qk\8j

u<sft-HUj5w)*Uj n+j<t#4& sfw)pf~9j58j<tq|;\t\sUo\l y\j H3wGn9sUtJl <*p?~9j<8j<tqB= pfn9wGj q)j<t\yJw)j(~9j xFp?wkpf 5q)w)j

RHUj5w)l~ .In+tMLFt\nxsfo\wp?lWT

l)oJw)j<wo\j ∑k P(PN

tk/∈ D)

qkj5t\yAHUj<wGl0j5q oJj

yN,R0 (PNt0

)qkj<tJy HUj5w)l

Y0

4#& sfl>x\w)j5n+j<wGlqkj5lGqklt oJ5w)n( oJj<l>t\sUoJl>sUtqkwGj<tqo\j~+j xFpfw)pf 5qkwGj

R;:l<.Ln9~ j5lGq uC\sUn+l)n l)o8l)pf8j<tq *fwkpft\yy5l ~+j

y\ oJq|;t#.Ln9tFoJjMxFpfl y\jMvp?t\n5w)jBlGn9*UtJnJLFu<p?qkn HUjBlGo\w ~_p u5sUt HUj<w)*fj<t\u5jf4n+t\l)n ;tJsUo\l y\<u5n9y\sUtJl y#.LoJq)n T~9n+l)j5wZoJty\sfvp?n9t\jmy\j>qCp?n9~+~+j L32Jj?4

iPNf

Page 161: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V )= 24/ 2:%) $98!$&% ) 2/1$&8 )=%:%) " ) ':0/1)=% !) *

') "&2><& sfo\lcu<sft\l)n+y\<wGsUt\lc~_pm87<8jZa bmVJdj5qc~9j5l 75j5l H?pf~+j<o\wGlcy\jZxFpfwkp? qkw)j5lo\jMxsfo\w ~9jZuCFp?x\n+q)w)jHjqMwkp?x\xj<~+sUt\l o\j ~9j>x\wGn 2y\j wG5=<wGj<t\u5j(j5lGq

7.1564

"!#%$&'(*),+

& sfo\l j j<u5q)o\sUt\l qksfoJq6y#.Ip sfw)yvy\j5l qkj<lWqklp HUj<u~9p p?l)j . $43e6sU8jBsUtv~/.Ip yJ ),(y\n q|;sftvy\<lGn9wGjqkj5lGq)j<wB~_p?H?pfwGn_prqkn9sftlGn9 o\~ qCp?t\<jmy\j

M, δj5qN4Factj :jq|;F~+j>y\sfvpfn+t\j

Dj5lGq=LJ23 oJt\j>=sUn9lxsUo\w

qksfoJqkjm,[40, 180]

j5q~9p(u5sUt\lGqkpftq)jCy(R)

j<lWq0LJ23>,100

4\e j<l H?pf~+j<o\wGlx\pfwkpfn+l)lGj<tql)o8lkp?8j<tq~_p?w)*Uj5l8xsfo\w8t\jxFpfl n_p?n9l)j5w~+j<l8w)5l)o\~ qCprqkl<4 a tj :jq|;co\t u|pf~+u<o\~Zy\jxJw)s p n9~+n qkyJjl)sfwGqkn+jzy\jy\sUpfn+t\jt\sUo\ln9t\yJno\j o\j~_p l)sUwWqkn9j y\j

[40, 180]X l)o\w~9j5lp?qkoJw)n+q)<ly#.IsfxJqkn+sUt\l o\jt\sUo\l

u<sft\l)n+y\<wGsUt\lC^ j<lWq6j 23q)w)7<8j5j5tq wkpfwGjMj5q L32Jj5wCy(R)

,100 = S0 = K

xFpfwkp?n+q qkw<l wkpfn+l)sUt\t\p ~9j?4& sfo\lB= pfn9lGsUt\l Hrpfw)n+j<w

Ny\j q)j<~+~9j>lGsUwGqkj oJj

N = [N0(√

2)j−1]sU

j = 1 . . .y\<lGn9*UtJj ~9j tJsU w)j

y\j0Hrpf~9j5o\w)l y\jN oJjB~ .IsUt-qkj5lGq)j X

[·] y\5l)n+*Ut\jZn+u<n3~_pxFp?wGqkn+j6j<tqkn5w)j|^4UactJl)o\n qkjf;ft\sUo\l = p?n9l)sft\l;Hrpfw)n+j<wδj5qM

yJj q)j<~+~9jml)sUwWqkjo\jδ = δ0√

2(j−1)αδ

j5qM = [M0

√2

(j−1)αM]4

& sfo\lx\wGj<t\sft\lN0 = 2 = M0

jqδ0 = 50

4 &MsUo\l L32Jsft\lo\t\j&H?p?~9j<oJw y\jαδ

xsUoJw~9po\j5~9~9j t\sfo\lqkj5lGq)sUt\ly\n :5w)j5tq)j<l H?pf~9j5o\w)l y\j

αM4 sUoJw6u \po\j H?pf~+j<o\w y\j

αδtJsUo\l6w)j5xsUwGq)sUt\l6~9jMx\w)n 28sP@Uj5t

y\sUtJt\8xFpfw ~/.Ipf~+*UsUwGn q)\8j X u<pf~9u5o\~9 j<t ~9pft)<pftq(N g =sUn9l>~ .Epf~+*Usfw)n qk\j1^;:~ .I5u|pfwWqGT qA@ xj8yJj8u<j8x\w)n 2 ;~_p u5sUoMHUj5wGqkoJw)jz8sP@Uj<tJt\jj<t

t0j5q ~/.L<u|p?wGqGT/q @ xjy\j~9p u5sUoMHUj5wGq)o\w)j?4 sUo\w yJsUt\t\j5w o\t\jn9y\5jy\j

~/.LsUwGy\w)j y\j*Uw)pft\y\j5o\w6y\j<lx\pfwkpf 5q)w)j5l<;3y\n+l)sft\l6xFpfw6j 23j<8x\~9j o\jmxsfo\wj = 10

;3u<j5~_p-lGn9*UtJnJLFj! oJjN = 64

j5qZlGo\nJHrpftq~9j5l H?p?~9j5o\w)lyJjαδ

j5qαM

sUty\sUt\tJj>y\pft\l~9j qkp ~+j|pfovu<n T y\j<lGl)sUoJlZ~+j<l H?pf~+j<o\wGly\ou<sfo\x\~9j

(δ,M)

αδ \ αMi iU4EN D D34EN H

g 4 h X h 4EDUN3; h K^ -. 0/2143652.7/98 X h 4ODfN3;ED g K Z^ X h 4EDUN3; iUiPN&ZUNU^ X h 4EDUN3; h NfNfH h ^i X iU4EN h ; h K^ X iU4EN h ;EH h DU^ - 612.436/2:<;9=98 X iU4EN h ; iUiPN&ZUNU^ X iU4EN h ; h NfNfH h ^iU4EN X g 4ED&ZJ; h K^ X g 4ED&Z3;EH h DU^ X g 4OD ZJ;ED g K Z^ ->:T6/2=3(17=2198 X g 4ED&ZJ; h NfNfH h ^D X g 4 g N3; h K^ X g 4 g N ;EH h DU^ X g 4 g N3;ED g K Z^ X g 4 g N3; iUiPN&ZUNU^ ->:T?:71430.2121257.98

a t*fwkpfl5; sUtvp(n9tJy\no\M~+j<l H?pf~9j5o\w)l u<sUwGw)j<lGxsUt\yFp?tqklpfo328l)j<oJn9~+l yJju5sUt-Hfj<w)*fj<t\u5jf4\a tj :jq|; xsUo\wuCFpo\jH?pf~+j<o\wy\j

αδsUt l5.Ep?xj<w)u5j!H wkp l)o\w8~9j5l8*Uw)pfx\\n(o\j<l8u5n T p?x\w 5l<;o#.In+~Zj23n9lWqkjo\t\jHrpf~9j5o\w

l)j5o\n9~ xsUoJwαM

j<tvy\j5l)l)sfo\l6y\j ~_p/ oJj<~9~+jZ~/.Ipf~9*fsUwGn q)\8jZl)j5 ~9jMtJj xFpfl u<sUt HUj<wG*Uj<w j5q6pfoy\j<lGl)o\l6y\j~_p/ oJj<~+~9j n9~ u<sft-HUj5w)*Uj?4 pfw j 2Jj5xJ~9j(xsUo\w

αδ = 0.6~+j(lGj<o\n+~ oJj ~ .IsUts lGj<w HUj t o\85w)no\j5j5tq

j<lWqαM = 1.5

y\sft\u sUtz=sUtJu<j ~9j>u5sUo\x\~+j(6.25, 362)

4csUn+u<n:~9j<lZw)5l)o\~ qCp?q)lBs q)j<to\lMxsUo\wB~9p Hrpfw)n9p?qkn+sUtl)n9 o\~ qCp?t\<jmy\j5lMxFpfw)pf 5qkwGj<l

N, δjqM

i hUg

Page 162: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

αδ = 0.6j5qαM

H?pfwGn9j` p x\w)sUxsfl)n+q)n9sft ' 4 i t\sUo\lvn+xsUlGj oJj

δq)j<t\yJjHUj5w)l

0xJ~9o\lKH n+q)j#o\j

h1246e6j5u<n n98xsUl)j

αδ >12

4 e.Lj<lGq Go\lGq)j~9ju|p?l-pIHUj5uαδ = 0.6

4` j8l)j5o\n9~cxsUoJwM

y\j8~_pzxJw)sUxsUlGn+qkn+sUt' 4 iU; n9u5npIHUj5u

d′ = 1X j5qMj5txJw)j<t\pftqM~_p?H?p?~9j<oJwB8n9t\n+8pf~ jmxsUo\w

β^;\y\sUtJt\j

αM > 2.64

j

Prix

105 156

7

8

6.5

7.5

1 1.5 2 2.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,1

t 1 PN,1

t 1 PN,2

t 1 PN,2

t 1 PN,3

t 1 PN,3

t 1

j

Std Pr

ix

105 15

1

2

3

4

5

6

7 1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105 15

0.1

0.2

0.3

0.4

0.5

0.6

1 1.5 2 2.5 Couverture BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 163: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Std Co

uvertu

re

105 15

1

0.5

1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1

M αM > 3 ! #"%$&(')*

j

Prix

105

6

7

8

9

6.5

7.5

8.5

1 1.5 2 2.5 3 3.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105

1

2

3

4

5

6

7

8

91 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

+

Page 164: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

j

Couver

ture

105

0.2

0.3

0.4

0.5

0.6

0.7

0.81 1.5 2 2.5 3 3.5 Couverture BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105

1

0.5

1.5

1 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1.5

M αM > 3.5

*

j

Prix

105

6

7

8

5.5

6.5

7.5

2 2.5 3 3.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 165: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Std Pr

ix

105

1

2

3

4

5

6

7

8

9

2 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

0.3

0.4

0.5

0.6

0.7

0.35

0.45

0.55

0.65

2 2.5 3 3.5 Couverture BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105

1

0.5

1.5

2 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

# % ' ' %$& ' ' ( $& # ) % αδ % αM > 1 ' % $ # '%) ' % # % ) αM # *

% % # )0 # ' * ( %

αδ = 1 )' $& αM = 1, 1.5 $ ) αM = 2

') αM ≥ 2.5

* ' & % $& & % * # ) &' " % # " %* % ' #& $ % ! ' # "%*

# $ # %' #% ' ) # ) '%) #%%$ '&')(' " % $ # * +&#$&$& %*-, !.

αδ = 1 '%) #%

0/

Page 166: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

t\jzlGj< ~+jxFpfl u<sUt HUj<wG*Uj<wu<~9pfn9wGj<8j<tq HUj5w)l~_p u5sUoMHUj<wWqko\wGjRZV ; pf~+sUw)l'o\jzxsUo\wαδ = 1.5

;uf.Lj<lWq~+jMu|p?l<43dBj<pfw o\sft\l&o\j ~9j5l q)\<sfw <8j5l 5qCp ~9n+lcyFpft\l ~+j<l xFpfwWqkn+j<l ' jq6'G' t\j u<sft\u<j5w)t\j5tq)pfpfn+l y\j<l>w)5l)o\~ qCp?q)lmy\j u5sUt-HUj5w)*Uj5t\u<jl)o\wmy\j<l>u<sfoMHUj<wWqko\wGj<l xsUt\uqko\j5~9~+j<lmvp?n9lMqksUo-GsUo\w)lmlGo\wy\j5ln+tq)<*Uw)pf~9j5l y\j ~_p u5sUoMHUj5wGqkoJw)jmlGo\w6~ .In+tq)j<w Hrpf~9~ j

[0, T ]43` .Ls lGj<w H?p?qkn+sUtvy\j u<sUt HUj<w)*fj<t\u5j>xsUoJw

y\j5l6u5sUoMHUj5wGqkoJw)j<l6xsUt\uqko\j5~9~+j<l l<.Lj23x\~+no\j u<j5wGqkpfn9tJj<8j<tq xFpfwhZN,R

tk= Etk(Y

N,Rtk+1

∆Wk)sUsUt

u<sfx\wGj<t\y o:.Io\tJj sUt\t\jzpfx\xJw)s|2Jn+vprqkn+sft y\jY N,R n9tJy\o\n+q-o\t\j sUt\tJjp?x\x\w)s|23n9p?qkn+sft y\j

ZN,R 4Z\5sUw)n( oJj<8j<tqMp HUj<u

αδ = 0.6;3sUty\j!H w)pfn+qu5sU8j5t\u<j5wZ,-u<sft-HUj5w)*Uj5w0HUj<wGlZ~9j3H3w)pfn x\w)n 2vxsUoJw

MlGo8lkpf88j<tq-*Uw)pft\y#4&M|pft\8sUn+t\l<;~_pH n+q)j<l)lGjvy\ju5sUt-HUj5w)*Uj5t\u<jvj<lGq(qkj5~9~+j<8j<tq ~9j5tq)j o\j

~/.LsUtztJj xj5oJqM~/.Ls l)j5w HUj5w p HUj<u oJttJsU w)j>y\j>qkj5lGqklBwkpfn+l)sUt\t\p ~9j?4Q t HUsfn+q y\sUt\u o\j~/.Ipf~+*UsUwGn qk\8j l)ju5sU8x sfwGqkjy\jpft\n5w)ju5sU\5w)j<tqkjp HUj<uu<j o\j~/.LsUt xj5oJqy\5y\o\n9wGjy\j5lj5lGqkn+vprqkn+sUt\l yJj~_pxFpfwWqkn9j'G'-j5q87<8jzxJ~9oJq fqn+j<o32 o\jxJw)!H o#4 a tj :jq|;c~+j<ll)j5o\n9~+l

αMj<t =sUt\uqkn9sft y\j

αδt\<u5j<l)lkp?n9wGj<lxsUo\w s l)j5w HUj<wo\tJjzu5sUt HUj<w)*fj<t\u5jy\j~/.Ipf~9*fsUwGn q)\8j

l)sftqmx\~9o\l pflo\j u<j5o32 x\w) H o\l>xFpfw~9px\w)sfx sfl)n qkn9sft '-4+i j5qmsUt u<sU8x\wGj<t\y vpf~ u<sU88j<tq~+j<ll)j5o\n9~+lBpfx\xFpfw)pfn+l)lkpftqBt oJ5w)no\j5j5tqBxj<oMHfj<tqM= pfn9wGj>u5sUt HUj<w)*fj<w X x\pfwBj 23j<8x\~9j|^~+jmqkj<wGj

CCy(R)2

hexp(− Mhβ+1

144Cy(R)2)N−1∑

k=0

E

(

N2

(hβ+1

2

3√

2, [P0,k+1]y − yN,Rk+1 (·), (PN,m

tk+1)1≤m≤M , (P

N,mtk+1

)1≤m≤M))

.

Q w<; n9~ pfx\x\pfwkp?n+qo\j ~ .Isftxsfo\w)wkp?n+qmj :j5u5q)o\j<w ,u<j5wGqCp?n9t\lmj<tJy\w)sUn qklmyJj 8j<n9~+~+j5o\w)j5lvp-GsUw)p?qkn+sft\l54 pfwZj23j5x\~+jf;\~9sUwGl o\j>yFpft\lB~9p x\pfwGqkn+j>'G'C;\sUt:HUj5oJqMvp-GsUw)j5w

P([AMk ]c)X y#.LsUzx\w)sIH3n+j<tqB~9jmq)j<w)8j

u<n TAy\j5l)lGo\lC^ ; sUts qkn+j<tq

P

( 1M

∑Mm=1 Um|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2

√1M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

>1

3h

β+12

∣∣∣∣(PN,m

tk+1, PN,m

tk+1)1≤m≤M

)

≤2 exp

(

−Mhβ+1 1

M

∑Mm=1 |gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2

18 1M

∑Mm=1

(|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2)2

)

.

sUo\wBpfw)w)pft\*Uj5wB~+j>pGsfwkpftqZy\j ~/.Ln9t\oFp?qkn+sUtu5n T y\j<lGl)o\l<;FsUt5u<w)n qMpf~9sUwGl

1

M

M∑

m=1

(|gj(PN,m

tk+1)|2 − |gj(PN,m

tk+1)|2)2

=1

M

M∑

m=1

(|gj(PN,m

tk+1)| + |gj(PN,m

tk+1)|)2(|gj(PN,m

tk+1)| − |gj(PN,m

tk+1)|)2

≤8Cy(R)2

M

M∑

m=1

(|gj(PN,m

tk+1)|2 + |gj(PN,m

tk+1)|2).

X ' 4LK\4EDU^

'A~ pfx\xFp?wkpfn qq)w)sUx =sUwWq>yJj vp-GsUwGj<w (|gj(PN,mtk+1

)| − |gj(PN,mtk+1

)|)2 xFpfw

4Cy(R)2 4#a tj j5q<;#5qkpftqy\sUtJt\ oJj

PN,mtk+1

= Tk(PN,mtk

,∆Wmk,k+1)

jqPN,mtk+1

= Tk(PN,mtk

,∆Wmk,k+1)

sU sft wkp?x\xj<~+~9ji h N

Page 167: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

o\jTk

y\<lGn9*UtJj~9pqkw)pft\lGn+qkn+sUt y\j~_puCFpfn+t\jy\j pfwGSfs HPN ;cn+~@ p y\jz=sUwWqkj<l8uCFpftJu<j<l o\j

gj(PN,mtk+1

)j5qgj(P

N,mtk+1

)lGsUn9j5tqMxJw)s uC\j<l54 n9tJl)n/;Jl)n

PN = XN j5q!o\j ~_p u<~9pfl)lGj G X o\nw)j5u<sUo H3wGj[P0,k+1]y − yN,Rk+1

^j<lWq oJt\j u5~_pflGl)j y\j>=sUt\uqkn9sft\lB`n9x\lGu Jn+q>X>sUtzs qkn+j<tq

|gj(PN,mtk+1

) − gj(PN,mtk+1

)| ≤ C|∆Wmk − ∆Wm

k |

j5q-sUt HfsUn+q oJj~ .IsUt *pf*ft\jo\t = pfu5q)j<o\wh4 a t *U<tJ<wkp?~/;u5sU8j8~9pu5~_pflGl)j

[P0,k+1]yt#.Lj<lWq xFp?l

o\t\j8u<~9pfl)lGj8y\j=sUt\uqkn9sft\l `n9x\lGu \n q>X[P0,k+1]y − yN,Rk+1

t:.EpxFp?l y\j8wkpfn+l)sUt y:.I7qkw)j8o\t\j8u<~9pfl)lGj8y\j=sUt\uqkn+sUt\lZ`n9x\lGu \n q>Xmj5qBy\sft\u G t\sftxJ~9o\l54 pfn+lZy\pft\lZt\sfqkwGj>u|pflZxFpfwWqkn9u5o\~9n+j5w6y\j>~_p pfl)j . $;Jn9~t#.Lj<lWqBxFp?ly\n 8u5n9~9jMy\jmlGj>u5sUt-Hrpfn9t\u5w)jo\jm~_p(p-GsUwkp?q)n+sUt X '-4IKJ4ODU^ j5lGqqksfoJqy\jm87<8j q)w)sUxv=sUwGq)jX HUsfn9w>5*pf~+j<8j<tq ~_pl)sUo\l TAlGj<u5q)n9sft K\4ED34IHzsf sUt l<.Ipfxj5w )5sUn+qo\j~9j<lm=sUt\uqkn9sft\l>j<lGq)n98<j5l t\jl)sUtqxFpflB~+sUn9ty#.L75qkwGj(`n9xJl)uC\n+qYX1^4

"!#%$&'(*),+

& sfo\l x\wGj<t\sft\l u<jq)qkjv=sfn9lαδ = 0.6

jq ~9p p?l)j . $ -%<3?: 8P4 csUn9u5n ~9j5l-w)<lGo\~+qkp?qkl-s qkj5t o\l xsUoJwy\n :5w)j<tqkj5l=H?p?~9j5o\w)lBy\j

αM4

j

Prix

105 15

6

7

8

5.5

6.5

7.5

8.5

1 1.5 2 2.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105 15

1

2

3

4

5

6

7

8 1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 168: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

j

Couver

ture

105 15

0.4

0.5

0.6

0.7

0.8

1 1.5 2 2.5 Couverture BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105 15

1

0.5

1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

"!#$%&'&)(&*,+&&&-.*,/&'0&&%12*%31*$45(&

αδ = 0.66+)+78&1:9;<>=0*,&*%@?A11*84B<C0D*

yN,RE1*,&)

F170GHI<>&JK&+*,=1,*L MHCM)GB17*7NOJP,..K=&+)&<>QR1*=&C

0JK(*+&>17*αM

NTSUQLV4B&<>1WXO,&<'0-=&.0*KαM > 2

NTYZ*U5(URL ;![\

αδ = 0.63X?],&*,(^01_U(*+&*_RL&(&*.*,&

t0N2`aG))L<>=&00*,=Fb?c11*,54B<C^.$d

yN,R"&, ^.K931*=&)L-&(*+&K'R

&(&*.*,N

egfihjflk m%noqp2rtsUuTvw xqyUp7zB|xqyUvEsu~ni

?A+*$.M<>U1_&&<>&)1*$4C9F@?c0,.t0<Cq=&+<C_31*4C9L.

0,.).tk17*

0 ≤ k ≤ NN2O,J).=*&,.)J%.*&*RC=\*<C\.0

1*$42WH&1* )jj = N

L\.031*,$4C&,=+)09%L\.15GNSj08jUF,U!8(&0&_1*.<C/\.*,&

N = 20WK0 = 20

&&q*,M = 5000 \L17*QP,&ZR&&*J>@?c+*,$M<>5VW2J.*&QJJK1*4"=&L1*

b?c0+*.M<>g\1*RQ*<'0gJ )|B M0& 5(J B43X?]=&*,R_B430,.

tj17*j = 4, 9, 14, 20 3*<3*,H?A,J^.*,&0&<>&)8W)@?c***_,*0g1*4>&,+*.17*_&_(&*_

S=&0(=& 17Q0<'R^.0j<'7=&&&\.Q&N

)

Page 169: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

S

Prix

80 90 100 110 12085 95 105 115

10

20

5

15

25

Prix algorithme Prix Black-Scholes

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,1

t 1 PN,1

t 1 PN,2

t 1 PN,2

t 1 PN,3

t 1 PN,3

t 1

. + !"t20 = T = 0.5

S

Prix

90 100 110 120 130

10

20

30

40

5

15

25

35

Prix algorithme Prix Black-Scholes

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,1

t 1 PN,1

t 1 PN,2

t 1 PN,2

t 1 PN,3

t 1 PN,3

t 1

#$%&(' N*),+-\.Q1*,$4t14 = 0.35

.

Page 170: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

S

Prix

80 90 100 110 12085 95 105 115 125

10

20

30

5

15

25

Prix algorithme Prix Black-Scholes

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,1

t 1 PN,1

t 1 PN,2

t 1 PN,2

t 1 PN,3

t 1 PN,3

t 1

. !"t9 = 0.22

S

Prix

90 100 11095 105 115

10

20

5

15

Prix algorithme Prix Black-Scholes

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,1

t 1 PN,1

t 1 PN,2

t 1 PN,2

t 1 PN,3

t 1 PN,3

t 1

# $%&(' N ' +!-Q1*4&t4 = 0.1

Page 171: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V )= 24/ 2:%)( / )& sfo\l pf~9~+sUtJl qkj<lWqkj5w ,x\w)5l)j<tq ~_p:H?pfwGn_p?q)n9sft l)n+ o\~+qkpft\5j8y\j<l-xFpfwkp? qkw)j5l

N;δj5qM

yFpft\l-~+ju|p?l yJj8~9p pfl)j . $ j5q xsfo\w ~ .Epf~+*Usfw)n qk\j x\wG<lGj<tq)yFp?t\l ~_pxFpfwGq)n9j ')'G'C;uCFpfxJn+qkwGjD ;xsfo\w ~+j<lacb>VJd w) F<uC\n+j<l<;:sf sUt x\wGj<t\y ~9j(pr23n9 o\,vuCFpo\j x\pfly\j(qkj<8x\lj5tq)w)j ~+j(w)5l)o\~ qCp?qy#.Lo\tx\wGs ~5j y\jw)<*fw)j<lGl)n+sUt jq(~ .Is lGqCp?u<~9j?4bmpft\l u<j8u<pfl<;u<sU88jn+t\y\n( oJyFpftJl ~_pxFp?wGqkn+j ')'G'C; ~+j<lqkJ<sUw<8j<l 'G'C4ED3;?'G'C4EHJ;?'G'C4LK>jq ')'C4EN l)sUtq j5t\u<sfw)j H?pf~9p ~+j<l yFp?t\l u5ju|p?l j5q y\sUtJu~9p x\w)sfxsUl)n qkn+sUt(' 4 i<*Upf~9j5j5tq<4#&MsUo\l>t\sUo\l p?qGqkj5t\y\sUt\l>y\sUtJu8,s lGj<w HUj<w ~9j 87<8j lGj<o\n+~ y\j u5sUt-HUj5w)*Uj5t\u<j=sft\u5q)n9sUty\j

αδxsUo\w

αM4

& sfo\lBu \sfn9l)n+l)lGsUt\ld = q = 1

4 ` .Iacb>VJdMd oJj t\sUo\l HUsUo\~+sUt\lBpfx\x\wGs u Jj<wBj<lGqM~_p l)o\n H?pftqkj

dStSt

= µdt+ σdWt,

dYt = (−rYt − θZt)dt− ZtdWt + dKt,

YT = (K − ST )+, Yt ≥ (K − St)+, 0 ≤ t ≤ 1,∫ T

0

Yt − (K − St)+dKt = 0

pIHfj<uθ = µ−r

σ

4#V n ~/.LsUt u<sU8xFpfwGj p HUj<u ~+j<lt\s?qCp?q)n9sUt\lMy\o e6\pfx\n+q)w)j D ; u5j5q>j 2Jj58x\~9j-u5sUw)wGj<l)xsft\y,f(t, x, y, z) = −ry − θz

jqg(t, x) = (K − ex)+

4e6j5u<n u<sfw)w)j5l)xsUtJy y\sft\u pfou|pflmy#.Lo\tx\o3qpf8<wGn9u<pfn+ty\jlGqkwGn9Sfj

Kj5qu5sUoMHUj5wGqZlkp?t\lu5sUtq)wkpfn+tqkj yJjmxsUwWqkj=j<o\n+~9~+j X y#.LsUo\ty\wGnJHUj<w

f~+n9t\<pfn+w)j

j<t(y, z)

^4`j<lBxFpfwkp? qkwGj<lByJj ~9p y3@3t\pfn(o\j y\j

SjqMy\j ~/.LsUxJqkn+sUtlGsUtqB~9j<lZl)o\n H?pftqkl

µ σ r T S0 Kg 4 g N g 4IK g 4 g N i i gfg i gfge6j5qMj23j5x\~+j j5lGqMq)n9wG>y\j Y R& g Kf] sf~9j5lMpfoJq)j<o\wGlMyJsUt\t\j5tq oJtxJw)n 2y\j w)5=5w)j5t\u<j

Y0 = 13.554

& sfo\loJqkn+~9n+lGsUt\l ~9p-87<8jxJw)s u<5y\o\wGj>xsUo\w qkj5lGqkj5wZ~9p&Hrpfw)n9p?qkn+sUtl)n+ oJ~+qCp?t\5jyJj<lZx\pfwkp? qkw)j5l o\jxsUo\wZ~9j>u<pflMyJj ~ .IsfxJqkn+sUtzp HUj<u yJn( <w)j5tq)n9j5~#y\j>qCpfo 2zy#.Ln9tqk<wG75q<4 αδ = 0.6

j

Prix

105

14

15

16

17

18

1 1.5 2 2.5 3 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 172: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

j

Std Pr

ix

105

10

5

1 1.52 2.53

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

-0.5

-0.4

-0.3

-0.55

-0.45

-0.35

-0.25 1 1.52 2.53

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8 1 1.52 2.53

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1.

j

Prix

105

14

15

16

17

13.5

14.5

15.5

16.5

17.5

1.5 2 2.5 3 3.5 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 173: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Std Pr

ix

105

10

5

15

1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

-0.5

-0.4

-0.45

-0.35

1.52 2.53 3.5

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1.5

* ' " % ) µ )% % 0.08

* ) %' % ' 0% ( % # * ) % )' % "$&(' % ' ' "*

+

Page 174: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e ' Zd aK\4% mdB' Z' Q &V '#` &ca bma V md BdMa V

j

Std Co

uvertu

re

105

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

1 2 3 4 5 6 7 8 9 10

-0.6

-0.5

-0.4

-0.3

-0.65

-0.55

-0.45

-0.35

-0.251.52 2.53 3.5

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 0.6

)' *µ = 0.08

*

j

Prix

105

13

14

15

16

17

18 1.5 2 2.5 3 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

-0.6

-0.5

-0.4

-0.55

-0.45

-0.35

1.52 2.53

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 175: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

Q ts l)j5w HUj <*pf~+j<8j<tqv~/.Lj23n+lGqkj5t\u<j yJj l)j5o\n9~+lvyJj u<sft-HUj5w)*Uj5t\u<jxsUo\wαM

j<t=sUtJu5qkn+sUty\j ~_pH?p?~9j<oJw y\j

αδ4 e6j<l l)j5o\n9~+l lGsUtq u5sU8xFpfwkp ~9j5l , u5j<o32s l)j5w HU5lxsfo\w ~+jBu|pfl acb>VJd"t\sUtw) F<uC\n+jf4

bjvx\~+o\l<;j5t u<sU8xFpfw)pftq(y\j5l?Hrpf~9j5o\w)l(y\jαδ

n9yJj<tqkno\j5l-xsUoJw-~+ju<pfl-w) F<uC\n j5q tJsUt3TAwG F5uC\n/;sUtl<.Ipfxj5w )<sfn+q!o\j ~_p u5sUt-HUj5w)*fj<t\u5j-yFpftJlM~9j>u|pflBw) F<uC\nj5lGq x\~+o\lB~+j<tq)j j5q!o\j ~/.LsUtp qkj<tJyFpft\u5j,8pfx\x\wGs u \j5w ~+j x\wGn 2pf85w)n9u<pfn+tzxFp?w pfo TAy\j5l)l)oJl<4 Q tu<sUt\lWqCp?q)j(5*pf~+j<8j<tqMj5t s l)j<w H?pftq~9j5lMu<pflαδ = 1.5

pIHUj5u~_p pflGj . $ jqαδ = 0.6

p HUj<u~9p p?l)j . $ -%<36: 8 oJj~/.Ipf~9*fsUwGn q)\8ju5sUw)wGn9*Uj~9j y\w)n = qMxsUo\w wGj5q)w)sUoMHUj5w ~9j xJw)n 2zw)n+l o\jT t\j<o3qkw)j?4#e6j5ly\j<o 2 j 2Jj58x\~9j5l8sUtqkwGj<tqM<*pf~+j<8j<tqo\j~9sfw)l o\j(~ .Ep?x\x\w)s|23n9p?q)n+sfty\j

Yj5lGqmlGo8lkpf88j<tq sUt\t\j?; ~ .Ep?x\x\w)s|23n9p?q)n+sftl)oJw~9p8u5sUoMHUj5wGqkoJw)j

j<lWq sUt\tJjf4Pact LFt#;Pj5t(u<sfxFp?wkpftq ~9j5l u|p?lαδ = 0.6

xsUoJw ~9j5l pflGj<l . $"jq . $ -%<3?:48sUt u5sUt\lGqkp?qkjo#.LsUt*p?*Ut\j j|pfo\u5sUo\x ,pGsfoJqkj5wBy\j5lMxsU~ @ t U8j<lZ~9s u<pfo32y\j y\j5*Uw)

1xsUo\wMpfxJx\w)s uC\j5w

Y4

iP[?K

Page 176: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

% 4

xJw <lp HUsUn+w HU<wGnJLF ~9p H?p?~9n+y\n qk>y\j5l w)5l)o\~ qCp?q)lMy\j u<sUt HUj<wG*Uj<tJu<j-y\sUt\t\5l yFpft\lM~_p xFpfwWqkn9j>')'C; t\sfo\lxsUoMHUsft\lMu<sfxFp?w)j5wB~+j<lZy\j<o32pf~9*fsUw)n q)\8j5l\u<j5~9o\n#y\j ~_p xFpfwGq)n9jm'6jqMu<j5~9o\n#y\j ~_p xFpfwGq)n9j>'G'C4 ` .Lo\tj<lWq pfl)l)oJw yJj<l n qk5wkp?q)n9sUtJl(y\j n9u<pfw)y , u Fp/o\jn9t\lWqCpftq y\jy\n+l)u<wG5q)n9lkprqkn9sft

tkjq uC\j<wGu Jj,

pfx\xJw)s uC\j<wZj<tz87<8jq)j<8x\lYtk

jqZtk

4\`j>l)j5u<sUtJyt:.Ep xFpfl j<lGsUn9ty#.In qk5wkp?q)n9sUtJl6y\jcn+u|p?w)yu|pfw~9jl)uC\<pj<lWq j 23x\~9n+u<n qkjf;jq(qkwGsUoMHUjxFpfw y\j5l-wG<*UwGj<l)lGn9sUtJl-lG<xFpfwG<j5l

YjqZ4& sfo\l-pf~9~+sUtJl#HUsUn9w

u<sf8j<tqBl)j u<sfxsUwWqkj5tqB~9j5lBy\j5o32pf~+*UsUwGn qk\j<ll)o\wBo\tj 2Jj5xJ~9j>qA@ xjf4& sfo\lt\sfo\lx\~9p)5sUt\l yFp?t\l~+j(u<pflyJj-~/.LsUx3qkn9sftj<o\wGsUx<j5t\t\j pIHUj5u-y\n( <wGj<tqkn+j<~ y\j qCpfo32y:.In+tq)<w)7q|;pIHfj<uzo\t\j pflGj+. $4 `jw)<lGo\~+qkp?qy\o qk\5sUw 5j'C4EHt\jxj<w)8j5q qkJ<sUwGno\j5j5tqxFpfl y\j= pfn+w)jqkj5t\y\wGj j5t75jqkj5x\l

N;MHUj5w)lM~ .In+tMLFtJn:j5q

δHUj<wGl

04 n9t\lGn#yFpft\lBo\tzx\w)j5n+j<w6qkj5x\l5;FtJsUo\l

pf~+~9sft\l LJ23j5wNjqδ; o\t y\sfvpfn+t\j

D = [40, 180]j5qM= pfn9wGj>qkj5t\y\w)j

MHUj<w)l ~/.Ln9tML\t\n/4\& sUo\lp?~9~+sUt\l

pf~+sUw)lu<sU8xFpfwGj<wZ~+j<lBu5sUt HUj<w)*fj<t\u5j<lyJj<lMy\j5o32p?~9*UsUwGn qkJ8j<l54 N = 5

;δ = 20

M

Prix

51. 10 52. 10 53. 10 54. 10 55. 106

7

8

6.5

7.5

Algo explicite Algo itérations de PicardPrix BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 177: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

M

Std Pr

ix

51. 10 52. 10 53. 10 54. 10 55. 10

0.1

0.2

0.3

0.4

0.5

Algo explicite Algo avec itérations de Picard

P0

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 N = 10 δ = 10

M

Prix

51. 10 52. 10 53. 10 54. 10 55. 106

7

8

6.5

7.5

Algo explicite Algo avec itérations de PicardPrix BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

M

Std Pr

ix

51. 10 52. 10 53. 10 54. 10 55. 10

0.1

0.2

0.3

0.4

0.5

Algo explicite Algo avec itérations de Picard

P0

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 N = 20 δ = 5

Page 178: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e! ' BdMa N34 e Q md 'WV Q & bma V%m` Q dB' a V

M

Prix

51. 10 52. 10 53. 10 54. 10 55. 106

7

8

6.5

7.5

Algo explicite Algo avec itérations de PicardPrix BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

M

Std Pr

ix

51. 10 52. 10 53. 10 54. 10 55. 10

0.1

0.2

0.3

0.4

0.5

Algo explicite Algo avec itérations de Picard

P0

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 N = 50 δ = 1

M

Prix

51. 10 52. 10 53. 10 54. 10 55. 106

7

8

6.5

7.5

Algo explicite Algo avec itérations de PicardPrix BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

M

Std Pr

ix

51. 10 52. 10 53. 10 54. 10 55. 10

0.1

0.2

0.3

0.4

0.5

Algo explicite Algo avec itérations de Picard

P0

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 179: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

Q txj<o3q6w)j5vp?w o\j5w oJj ~9j5l pf~+*UsUwGn qk\j<l l)jMu<sU8xsUwGq)j<tq j5l)l)j5tq)n9j5~9~9j5j5tq y\jM~9p 87<8jZ= p)5sUtv,u<j5u<nFx\w<l oJjM~ .Ep?~9*UsfwGn qkJ8jpIHUj5uMn qk5wkp?q)n9sUtJl y\j cn+u|pfwGy qkj5t\y,>y\sUt\tJj<wco\tJjBx\~9o\l xjqkn+q)j0H?pfwGn_pftJu<jo\j~ .Ep?~9*Usfw)n qkJ8jj 23x\~9n+u<n qkjf4 Q t xj5oJq xj<t\lGj<w o\j~9pu|pfo\lGjj<lWq ~_pn+t\n+8n lkp?q)n+sft lGn9 o\~+qkpftJ<jj<t

(Y, Z)~9pzxFpfwWqkn9jj5t

Z∆Wxj5oJq GsUo\j5w(~+j8wU~9jy\j H?pfw)n9p ~+jy\j8u<sUtqkw f~9jf4 Q t xj5oJq-j5tMLFt

w)j5vp?w o\j<w o\j~+j n_p?n9l j<tqkw)j~9jzx\wGn 2 y\sft\t\x\pfw8~ .Epf~+*Usfw)n qkJ8jj5q~+jzx\w)n 2 yJjwG5=5w)j<tJu<j j5lGq~95* 5w)j<8j<tqBx\~9o\l= pfn ~9jyFpft\lB~9j>u<pflBy\j ~/.Ipf~9*fsUwGn q)\8jmj 23x\~9n+u<n+q)jf4& sfo\l p?~9~9sft\l(, x\w)5l)j5tq j<l)l)p1@Uj5w y\jv= pfn+w)j H?pfw)n+j<w(l)n+ o\~+qkpft\58j<tq(~+j<l-x\pfwkpf 5q)w)j5l-yFpft\l-~/.Ipf~9*fs?Tw)n qk\8j p HUj<un qk<w)p?qkn+sUt\lcy\j n+u|pfwGy#; xsUoJw HUsfn9w l)n~9j<l wG<lGo\~+qkp?qkl l)sftq6n9y\j5tqkn(o\j<l6,(u<j5o32s l)j5w HU5lxsUo\wZ~/.Ipf~9*fsUwGn q)\8jmj 23x\~9n+u<n qkjf4 αδ = 0.6

jqαM

H?p?w)n+jf4

j

Prix

105 15

2

3

4

5

6

7

1 1.5 2 2.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105 15

10

20

5

15

25 1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105 15

-0.1

0.1

0.2

0.3

0.4

0.5

0.6

1 1.5 2 2.5 Couverture BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 180: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e! ' BdMa N34 e Q md 'WV Q & bma V%m` Q dB' a V

j

Std Co

uvertu

re

105 15

1

0.5

1 1.52 2.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1

j

Prix

1051

2

3

4

5

6

7

8

9

1 1.5 2 2.5 3 3.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105

10

20

5

15

1 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

0.1

0.2

0.3

0.4

0.5

0.6

1 1.5 2 2.5 3 3.5 Couverture BSP0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 181: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Std Co

uvertu

re

105

1

2

0.5

1.5

1 1.52 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 αδ = 1.5

j

Prix

105

1

2

3

4

5

6

7

2 2.5 3 3.5 Prix BS

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105

10

20

5

15

2 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

0.1

0.2

0.3

0.4

0.5

0.6

2 2.5 3 3.5 Couverture BS

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

Page 182: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e! ' BdMa N34 e Q md 'WV Q & bma V%m` Q dB' a V

j

Std Co

uvertu

re

105

1

0.5

2 2.53 3.5

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 # '$ #0% ' '% # % ) ' "$& ! ' #* )' %

#' αM ' # αδ % $ # )

0 "$& $ &') $ $& * $ %$ % $& "% * #"%$& ) ' # ' $ $& ! #"%$&(% %' *

Page 183: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

iZD

Page 184: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

& sfo\l>pIHUsft\lmy\ ),qkj5lGqk ~_pKHrpfw)n9p?qkn+sUtl)n+ o\~+qkpft\5j(y\j<lmxFpfw)pf qkwGj<lN;δ; jq

MyFpft\lm~+j-u|p?lmy\j

~_p p?l)j3. $ xsUo\wBo\t\j(acb>VJd w) F<uC\n+jf4& sUoJl pf~9~+sUt\lyFpft\lMu5j(uCFpfxJn+qkwGj>qkw)pfn+q)j<wZx\~+o\lMj5ty\5qkpfn9~~_pH?p?~9sUwGn9lkprqkn+sft jq-u<sUo HUj<wWqko\w)jvy#.LsUx3qkn9sft\l(pf8<w)n+u|pfn+t\j5l<4 Q oJqkwGj~/.Ipf~+*Usfw)n qk\jx\wG<l)j5tq)yFp?t\l(~_pxFpfwWqkn+j 'G'G'C;ruCFpfx\n qkwGj6D3;PsU(~/.LsUt x\w)j5t\y ,Mu Fp/o\j6n9tJlGqCp?tq yJj y\n9lGu<wG5qkn+lkp?q)n9sUt

tk;P~+j pr23n9 o\ j5tq)w)j

~/.Ls lWqCpfu5~9j6jq ~_pl)sf~9oJq)n9sUt-y\o x\w)s ~5j6y\jsfn9t\yJw)j<l u<pfw)wG<l<;fsUt xj<oJq n9pf*Un t\j<w 5*pf~9j58j<tq yJj<o32pfoJq)w)j5lMqk\s yJj<lMxsfo\wMpfx\x\wGs u Jj<wM~9p l)sU~+oJqkn+sUty#.IoJt\j(a bmVJd 4 & sUo\lMpf~9~+sUt\lZ,x\wG<lGj<tq y\5qkpfn9~+~+j<wu<j5lMy\j<o 2pfoJq)w)j5lM85q)\s y\j<l

W = 6

tJj 85q)\s y\j yJj xJw)j<o HUj(y\j ~ .Ij 2Jn+lGq)j<t\u5j jqMy\j ~/.Lo\t\n9u5n+q)>,~ .Eacb>V3d wG F5u \n+j

Yt = g(T,XT ) +

∫ T

t

f(s,Xs, Ys, Zs)ds+KT −Kt −∫ T

t

ZsdWs,

Yt ≥ g(t,Xt),

∫ T

0

(Yt − g(t,Xt))dKt = 0X ' 4 h 4 iP^

u<sft\l)n+lGqkj , y\!LFtJn9wBoJt\j lGo\n+q)j y#.Iacb>VJd tJsUtw) F<uC\n+j<l

Y nt = g(T,XT ) +

∫ T

t

f(s,Xs, Yns , Z

ns )ds+ n

∫ T

t

(Y ns − g(s,Xs))

−ds−∫ T

t

Zns dWsX ' 4 h 4EDU^

y\j l)sU~+oJqkn+sUt(Y n, Zn)

4 Q ty\!LFtJn+q>pf~9sUwGlKnt = n

∫ t

0(Y n

s − g(s,Xs))−ds

;:jq>sUt8sUtq)w)j X HUsUn9wY a + U[r]_^o\j~9sUwGl o\j n qkj<tJy4HUj5w)l~/.Ln9t LFt\n/; (Y n, Zn, Kn)

qkj<tJy4HUj5w)l(Y, Z,K)

l)sU~+oJqkn+sUty\jX '-4 h 4+iP^ 4` .In+y\<jj<lWqpf~9sUwGl8xsUo\w

nLJ23f; y#.Ipfx\x\wGs u \j5w~9p l)sU~+oJqkn+sUt y\j ~ .IoFp?qkn+sUt X ' 4 h 4EDU^8, ~ .Ep?n9y\jzy\j

~/.Ipf~+*Usfw)n qk\8jzj 2Jxsfl) yFpftJlv~_p xFp?wGqkn+j')'C4 a tj j5q|; X ' 4 h 4EDU^j<lWqo\t\j acb>VJd pIHUj5u o\ty\w)n HUj<w`n9x\lGuC\n+qYXf4 dBj<pfw oJsUt\l o\j H3o~9j<l8w)<lGo\~+qkp?qkly\j Y a + [P] ; Yn Y

jq oJj ~ .Isft x j5oJq

Page 185: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

l<.Ip?qGqkj<tJy\w)jp HUj<uu<j5qGqkjqk\s y\j,q)w)sUo HUj<w y\j5l sUw)tJj<l n9tJ=5w)n9j5o\w)j5l X y\jx\~+o\l j5t x\~9o\l>x\wG<u<n+l)j5loFpft\y

n→ ∞ ^6y\ox\w)n 2zpf8<wGn9u<pfn+t#4

W 6

tJj p?oJqkwGj(n9yJ<j u<sUt\lGn9lWqkj-,o3qkn9~+n+l)j5wB~9j5l wG<l)oJ~+qCprqkl yJj Y RZe a g Dr]/4`j<lpfoJqkj5o\w)l u<sft\l)n+y<wGj<tqm~9ju|p?lMsf~/.Ls lWqCpfu5~9j

g(t,Xt)j<lWq o\t\j l)j5n+vp?wGqkn t\*p?~+ju|,fy\~9,f*yJj ~9p =sfw)8j

g(t,Xt) = g(0, X0) +

∫ t

0

Usds+

∫ t

0

VsdWs + At

pIHfj<uU

j5qV

y\j<l x\wGs u<j<lGl)o\l xJw)!H n+l)n ~9j<lGHf<w)n L pftq ~9p u<sUtJy\n+q)n9sUt y:.In+tqk5*Uwkp n9~+n qkE∫ T

0|Vt|2 + |Ut|2dt < ∞ jq

Ao\t xJw)s u<j5l)l)oJlvu|,fy\~9,f*\; pfyFp?xJqk?;cu<wGsUn9lGlkpftq|; j5qKHf<w)n L pftq~9p

u<sft\y\n+q)n9sft y:.In+tq)<*Uw)p n9~+n qkE|AT | < ∞ 4 bmpft\l u<ju|p?l<; sUt lkpfn q o\jzuC\j5w)uC\j<wo\t\jl)sf~9oJq)n9sUt ,

X '-4 h 4+iP^6w)j H n9j<tq ,uC\j5w)uC\j<wMo\tzqkw)n+x\~9jq(Y, Z, α)

HU<wGnJL pftqM~ .Eacb>V3d

Yt =g(T,XT ) +

∫ T

t

f(s,Xs, Ys, Zs) + αs1Ys=g(s,Xs)[f(s,Xs, g(s,Xs), Vs) + Us]−ds

−∫ T

t

ZsdWs,

Yt ≥g(t,Xt)X ' 4 h 4IH^

pIHfj<uE∫ T

0|Yt|2 + |Zt|2 + |αt|2dt

4a t u<sU8xFp?wkpftqp HUj<u X 'G'G'C4OD 4+iP^ ; sUt u5sUt\lWqCp?q)j oJj-u5j<~_pw)j!H n+j<tq,8j23x\~+sUn+q)j<wB~9j>= pfn+q o\j-~ .Isftl)pfn+qy\ ),J;Jj<tzoJq)n9~+n+lkpftq6~_p-wGj<pfw o\j-4IHJ;o\j>~9jx\w)s u5j<lGl)o\lMu5w)sUn+l)l)pftq

Ky\j>~9p-l)sU~+oJqkn+sUtyJj>~/.Iacb>VJd

X '-4 h 4+iP^6l)j5wkpy\j ~_p-=sUw)8j

dKt = αt1Yt=g(t,Xt)[f(t,Xt, g(t,Xt), Vt) + Ut]−dt,

pIHfj<u0 ≤ αt ≤ 1

4r` .In+t\u<sUtJt o\j xsUoJw y\qkj<wGn+t\j<wKy\j H3n+j<tq yJsUt\u ~9j x\wGs u<j5l)l)oJl

α4 sUo\w 8sUtq)w)j5w

~/.Lj23n+lGqkj5t\u<jjq ~ .IoJt\n9u5n+q) yJj~/.L oFprqkn9sftu5n T y\j<lGl)o\l5; ~+j<l pfo3qkj<oJw)l oJq)n9~9n l)j<tq>o\t\jwG<*UoJ~_pfwGn9lkprqkn+sft y\jX '-4 h 4EH^ 4Fa txFp?wGqkn+u<oJ~9n+j<w5;JoJt\j =sfn9lu JsUn9lGnoJtxFpfwkp? qkw)jmy\j w)5*Uo\~9pfw)n+lkp?q)n+sUt

n;\n9~+ln9tqkw)s y\oJn9l)j5tq

o\t\j l)oJn+qkj>y\j>=sUt\uqkn9sft\lφn ∈ C∞ qkj5~9~9j5l o\j

0 ≤ φn ≤ 1j5q

φn(x) =1l)n |x| ≤ 1

2n,

φn(x) =0l)n |x| ≥ 2

2n.

xFpfwWqkn9w y\j>u<jq)qkj>lGo\n+q)j>yJjm=sUt\uqkn+sUt\l<;3~9j5lBp?oJqkj5o\w)lZy\ LFt\n+l)l)j5tqMoJt\j>l)o\n qkjmy#.Iacb>VJd xFpfw)pfqkwG<jxFpfw

n

iZfK

Page 186: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e c' BdMa h 4 Q Z' Q &mV%cdM'We ' &a V

Y nt =g(T,XT ) +

∫ T

t

f(s,Xs, Yns , Z

ns ) + φn(Y

ns − g(s,Xs))[f(s,Xs, g(s,Xs), Vs) + Us]

−ds

−∫ T

t

Zns dWs.

X ' 4 h 4LK^

a t u5sU8xFpfwkp?tq p HUj<u X ' 4 h 4EHU^; sftHUsUn+q oJj(u<j<~9p8wGj!H n9j5tq,8wG<*UoJ~_pfwGn9lGj<wB~+j>qkj5w)8jαs1Ys=g(s,Xs)

4`j<lcpfoJqkj5o\w)l sftqkwGj<tq p?~9sUwGl o\jB~_pl)sU~+oJqkn+sUt y\jZu<jq)qkjMacb>V3du5sUt-HUj5w)*UjB~+sUw)l oJj

n→ ∞ HUj5w)lc~_pl)sf~9oJq)n9sUty\j X '-4 h 4EH^ 4Fa tsUoJqkwGjf;\sUtp

Yn Y4F`c.Iacb>VJd wG<*Uo\~9pfwGn9l)5j X ' 4 h 4LK^6j<lWq oJt\j acb>VJd

,-y\wGnJHUj5w`n9x\lGu Jn+q>X?4 Q txj<o3qZy\sUtJum~ .EpfxJx\w)s uC\j<w6,-~/.Ipfn+y\jMy\jm~ .Ep?~9*Usfw)n qkJ8jMy\<u5w)n qyFpftJl~_p xFpfwGq)n9j'G'6j5qMj<lGx<w)j5w s q)j<t\n+wBy\j5l sfw)t\j5l lGo\x<wGn9j5o\w)j5l X y\j x\~9o\lBj5tx\~+o\lBx\w)5u<n+l)j<lM~9sUwGl o\j

n → ∞ ^y\ox\wGn 2zpf8<wGn9u<pfn+t#4

)= $K) !)=% :< =

e6j<l y\j<o 2 85q)\s y\j<lZT6x<t\pf~9n+lkp?q)n9sftjqwG<*Uo\~9pfw)n+lkp?q)n+sUt T u5sUt\y\o\n+l)j5tqm,pfx\x\wGs u Jj<w t\sft xFpflMo\t\jacb>VJd pfn+lMo\t\j-lGo\n+q)j(y#.Eacb>V3dx\pfwkpf85q)w)5j(l)sUn qx\pfw ~9j u<s j u5n9j5tqmy\j-x5tFpf~+n9l)p?qkn+sUtl)sUn q xFpfw~9j8u5s j 8u5n9j<tq y\jwG<*Uo\~9pfw)n+lkp?q)n+sft#4 bjx\~9oJl<; ~_p8qk\s y\j8y\jvx5tFpf~+n9l)p?qkn+sUt p?x\x\w)s uC\j

YxFpfw j<t

y\j5l)l)sfo\l j5qM~9p 85q)\s y\j yJj wG<*UoJ~_pfwGn9lkprqkn+sftzpfxJx\w)s uC\jYxFpfwMp?o3TAyJj<l)lGo\l<4

& sfo\lmu<sUtJl)n9yJ<w)sft\lmxsUo\wmu5j-qkj<lWq>~9j-u<pflmy\ox\oJq>pf8<w)n+u|p?n9ty\j ~_plGj<u5q)n9sUt K\4IHJ;:pIHUj5u ~+j<l75j5lH?p?~9j<oJw)lBy\j xFpfw)pf 5qkwGj<l54bmpftJl(oJt x\w)j5n+j<wmqkj58x\l<;t\sUo\l(p?~9~+sUt\l LJ23j<w ~9j5l xFpfwkp? qkw)j5l y\j8w)5*Uo\~9pfw)n+lkp?q)n+sUtjq(y\j8x5tFpf~+n Tlkprqkn9sft#; x\o\n9l-= p?n9w)j Hrpfw)n+j<w-l)n9 o\~ qCp?t\<8j5tq

N; ~+jt\sf w)jy\jv=sUt\u5q)n9sUtJl y\j pfl)jj5q

M4 &MsUo\l

j<lGx<wGsUt\l-p?n9t\lGn;HU<wGnJLFj5w o\t j<t\u<pfy\w)j58j<tq y\jY0

x\pfw ~+j<l>y\j<o 2 85qkJs y\j<l j5q#HU<w)n LFj<wo\j~/.Lj<t3Tu|p?y\w)j5j5tq-y\j H n9j<tq-x\~9oJl(x\wG<u5n9l-pfo =o\w(jq ,j5l)o\wGj o\j~9j5l(u5s j 8u<n+j<tqkl-y\jx<t\pf~9n+lkp?q)n9sft jqw)5*Uo\~9pfw)n+lkp?q)n+sUtpfo\*U8j<tqkj5tq|4

& !#(

sUo\wy\n( <wGj<tqkj<l H?pf~+j<oJw)lmy\jn; t\sUo\lmoJq)n9~+n+l)sUt\l ~9p pfl)j . $ -%<3?:48 p HUj<u

αδ = 0.6jqαM = 2

4& sfo\lmpIHUsUtJlmj5tj :jq H3o yFpftJlm~+j<l u Fp?x\n+q)w)j5lx\w)5u<5y\j<tqkl oJj u<j-uC\sUn 2y\j-xFpfwkp? qkw)j(pfl)lGo\wkpfn q~_p u5sUt-HUj5w)*Uj5t\u5j-y\o n9pfn9lj5q y\j ~_p H?pfwGn_pftJu<jf4 &MsUo\lM8sUtq)w)sUtJlB~9j5lBw)<lGo\~+qkp?qklBy\j u5sUt-HUj5w)*Uj5t\u<j(y\ox\wGn 2 ;\y\j ~_p u<sUoMHUj5wGq)o\w)j j5qMyJj ~+j<o\wZ5u|pfwWqGT/q @ xj xsfo\w

nH?p?~_pftq

0.5, 1, 5, 104

iZN

Page 187: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Prix

105 15

20

15

25

0.5 1 5 10 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105 15

-0.7

-0.6

-0.5

-0.4

-0.65

-0.55

-0.45

-0.35

0.51 5 10

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105 15

10

20

5

15

0.51 5 10

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105 15

1

0.5

0.51 5 10

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 "$(5".K* .#%

nS=!><"$# 8 I% @?*"W%-".**"$ *%S * *"$ +*@

[

Page 188: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e c' BdMa h 4 Q Z' Q &mV%cdM'We ' &a V

pft\n5w)j y\sUtq sft p= pfn+q H?pfwGn9j5wN;δj5qM4 Q t lkpfn q oJj~ .Isfty\sUn+q>t\sUw)pf~+j<8j5tq pfx\x\wGs u Jj<w

~9jvx\wGn 2 y\jwG5=5w)j<tJu<jzxFpfw-j<t y\j5l)l)sfo\l<4 e.Lj<lWq n+j<t u<j o\j~ .IsUt s l)j5w HUjzxsUoJwn = 0.5

sUo14

pfwu<sUtqkwGjf;:xsUo\wn = 5

sfon = 10

;:~+j<ly\j<wGt\n5w)j5l H?p?~9j<oJw)lmy\jjq)j<lGq)<j5l x\wG<l)j5tq)j<tq qksfoGsUo\wGl

o\t x\w)n 2~95* <wGj<8j<tq lGo\x<wGn9j<oJw(pfo x\wGn 2y\jw)=<wGj<t\u5jf4 Q t xj5oJq(xj5t\l)j5w o\j8~/.Ipf~+*Usfw)n qk\8j-t#.EpxFpfl j<t\u5sUw)ju5sUt-HUj5w)*UxsUo\w u5j<l9H?pf~9j5o\w)l jq o\jl)nc~/.LsUtqkj<lWqCpfn q y\j<l9H?pf~+j<o\wGl(yJj

jlGo\x<wGn9j<oJw)j<l5;

sUtw)j<x\pfl)lGj<wkp?n+q j5ty\j<lGl)sUo\l y\ox\w)n 2 y\j wG5=5w)j<tJu<jf4e.Lj<lWq <*pf~+j<8j<tq>u5sU\5w)j<tq y#.Ls lGj<w HUj<w o\t\ju<sft-HUj5w)*Uj5t\u<j-8sUn9tJlBw)pfx\n+y\j yFpftJl ~+j u|pflMy\j

n*fwkpft\yu|pfwB~9j u5s j 8u<n+j<tqy\j(`n9xJl)uC\n+qYX y\oy\w)n HUj<w

n9tqkj5w H n9j5tq yFp?t\l ~+j<l>j<lGq)n9p?qkn+sft\lmy#.Ij5w)wGj<o\wGl o\j~/.LsUt pzs qkj5t o#4 ~9o\l>u<j5~9o\n TAu<n j5lGq(*fwkpft\y:;x\~+o\l~9j5lBu5sUt\lGqkpftqkj<lMyFp?t\lM~+j<l sUw)t\j5lMy#.Lj<wGw)j5o\wMl)sftqMn+xsUwWqCpftqkj5l<4

& !##

sUo\wZpfx\x\~+no\j5wu<jq)qkj>qk\s yJjf;Jt\sUo\lZy\j HUsUt\lBl)x<u5nJL\j<wB~9p-=sft\u5q)n9sUtφn4F&MsUo\lZ8s y\nJL\sUt\l6~9<*T

w)j5j5tq ~_p>yJ!LFt\n qkn+sUt oJn j<tj<lWq y\sUt\t\5j y\pft\l Y RZe a g Dr] ;qCpftq y\sUt\t\!o\jMt\sUo\l tJjMx\w)j5t\sUt\lxFpfl

φn C∞

φn(x) =1l)n |x| ≤ 1

2n,

φn(x) =0l)n |x| ≥ 2

2n

j5qφn

w)j5~9n_p?tqZ~+n9tJ|pfn+wGj<8j<tq0,

1l)o\wB~9j5lBn+tqkj5w H?p?~9~+j<l

[− 2n,− 1

n]j5q

[ 1n, 2n]4

a t sfoJqkwGjf; ~/.Ln9y\5j y\j ~9p 85q)\s y\j y\<u5w)n+q)j yFp?t\l Y RZe a g Dr]mj5lGqy\j y\sUtJt\j<wy\j5lpGsfwkpftqkly\j

Y4 'A~n+8xsUwGqkjy\sUt\uy\j*pfwkp?tq)n9w oJjz~_p y\n+l)u5w)5q)n9l)p?qkn+sUt y\j X '-4 h 4IK^&Hrp n9j5t =sfo\w)t\n+w-y\j<l

pfx\xJw)s123n+vprqkn+sUtJlxFpfwBpfo3T y\j<lGl)o\l54 sUo\wBu5j5qGqkj w)pfn9lGsUt#;JsUtH?py\n+l)u<wG5q)n9l)j5w X ' 4 h 4IK^yJj ~_p vp?t\n5w)jl)oJnJH?p?tqkj

Y Ntk

=Etk

(

Y Ntk+1

+ hf(tk, XNtk, Y N

tk+1, ZN

tk)

+ hφn(YNtk+1

− g(tk+1, XNtk+1

))[f(tk, XNtk, g(tk+1, X

Ntk+1

), V Ntk

)

+g(tk+1, X

Ntk+1

) − g(tk, XNtk

)

h]−)

,

hZNtk

=Etk(YNtk+1

∆Wk),

hV Ntk

=Etk(g(tk+1, SNtk+1

)∆Wk).

` .In+y\<j j<lWq o#.Lj<t oJqkn+~9n+l)pftqmo\t p?tFpf~+sU*Uo\j y\oqk\5sUw 5j y\j8u5sU8xFpfw)pfn9lGsUtxsUo\w ~+j<l acb>V3d y\n+lWTu<w5q)j<l<;UsUt xj<o3q = pfu5n9~9j58j<tq 8sUtqkw)j5w o#.Isft8pqksUo3qkj<l ~9j<l u Fp?t\u<j5l oJjZ~/.Ipfx\x\wGs123n9prqkn sUt-y\ju<jq)qkjy\n+l)u<wG5q)n9lkprqkn9sfty\sUt\t\j y\j5lMx\w)n 2l)o\x5w)n+j<o\wGl ,~ .Is lGqCp?u<~9jmj5qMyJsUt\u l)o\x5w)n+j<o\wM,

Y4

e6n T y\j<lGl)sUoJl<; tJsUo\lBx\w)5l)j<tqksUtJl ~9j5lBwG<l)oJ~+qCprqklBxsfo\wBy\n :5w)j<tqkj5l=Hrpf~9j5o\w)lBy\jn4

iZ[

Page 189: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

j

Std Co

uvertu

re

105 15

1

0.5

0.51 5 10

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Pr

ix

105 15

10

5

1 2 10 100

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105 15

-0.6

-0.5

-0.4

-0.3

-0.55

-0.45

-0.35

1 2 10 100

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Std Co

uvertu

re

105 15

0.1

0.2

0.3

0.4

0.5

0.6

0.71 2 10 100

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1.% %- 13 <<"$K '6#!* <".#&8,(54*"$#O# <U" .$*N 134".%)*%$"$

!<".#&8 ".: ". ) 1F! ".#&%$'() *%J!<".#&8 ()4".#!O/#! .)"$.$*"$"$5"$.n*"$#&%ON

E"V%-". $#n*.%%-".< "-/

n = 100 S@ "$4* !"$#-/%$#;213#!M%-*".K#!*%<UV*%! <"$#&8?*"$[

Page 190: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e c' BdMa h 4 Q Z' Q &mV%cdM'We ' &a V

~9jo\j<~~ .Epf~+*Usfw)n qk\jmu<sft-HUj5w)*Uj j5lGq n+tJ=<wGn9j5o\wZpfox\wGn 2zpf8<wGn9u<pfn+t#4

)= $K) !)=% #"#/;!" C*!)=

& !#(

& sfo\lzp HUsUt\ly\ ), s lGj<w HU~9j5lx\wGsUx\w)n+5q)<ly\j u<sUt HUj<wG*Uj<tJu<jy\j~9p 85qkJs y\j y\j x<tFpf~+n9l)p?qkn+sUt, x<t\pf~9n qkLJ23j X

n = 0.5, 1, 5, 10^4 tJj n9y\5j tFp?q)o\w)j5~9~9jj<lGqy#.Ep?x\x\~9n(o\j<w~/.Ipf~9*fsUwGn q)\8jp HUj<u

o\t\j x<tFp?~9n+q) Hrpfw)n9p ~+jf;o\nHrpfw)n+j<w)pfn+qZj5t87<8jmqkj5xJl oJjN;δj5qM

yFpftJlM~+j>u|pflBy\j ~_p pfl)j. $ -%<3?: 8P4- sUo\wq)j<lWqkj<w u<j5qGqkj H?pfw)n9p?qkn+sUt l)n+ oJ~+qCp?t\5jf;1sUt-xsUlGj

n = n0(√

2)(j−1)αnxsUoJw

n0 = 0.5j5qMsftzqkj5lGqkj ~_p u<sUt HUj<wG*Uj<t\u5j(y\j ~/.Ipf~9*fsUwGn q)\8jmxsfo\wBy\n :5w)j5tqkj5l=Hrpf~9j5o\w)lByJjαn4

j

Prix

105 15

10

20

30

40

50

60

70

80

90 0.1 0.3 0.5 0.7 1 1.5 1.8 2 2.5 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

αn !"

αn = 2"

2.5

" αn = 1.8

! "# #" $%" &'( ) * , + , -, .0/1. ""2& 34 " + " 5762." 7% 8 9%"

αn"! " " " : ;"<" : ; )

=> ? @%" αn = 1.8, 2, 2.5

j

Prix

105 15

12

13

14

15

16

17

18

0.1 0.3 0.5 0.7 1 1.5 1.8 2 2.5 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 K(*

αn = 0.111*0*1l0J&*g(*+F*10&<>&)(&*,UL(&*g0=\

*0&*,J1*4<C=*080 &HX&gM=*= =JHQ%<C=\.MHJ17=&0.011*H MQ1*B&,g01*4<>=&*08<C,.J(^*j=(0.*,117 *.1<>&

.

Page 191: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

HUj<wGl ~+j xJw)n 2pf8<wGn9u<pfn+t#4:`j<l#H?p?~9j5o\w)lαn ≥ 0.7

lGj< ~+j<tqmqkw)sUx =sUwGqkj5l>u|pfw>sUtl)jwGj5q)w)sUoMHUj8p?o3Ty\j5l)l)oJlcy\o x\wGn 2-pf8<wGn9u|p?n9t u<j o\n3j<lWq xj<o u5sU\5w)j<tqcpIHUj5u~9j<l w)<lGo\~+qkp?qkl prq)qkj5t\y\o\l y\j~9p 85qkJs y\jf4dBj<lGq)j<tq(~9j5l#H?p?~9j<oJw)l

αn = 0.3sUo

0.5 oJncy\sft\t\j5tq(~+j<l>w)<lGo\~+qkp?qkl>~9j<l>x\~+o\l u5sU\<wGj<tqkl<4`j<l w)T

l)oJ~+qCprqklBlGo\wB~_p u5sUoMHUj5wGqkoJw)j(u5sUtMLFwGj5tqMu5j<lMs lGj<w Hrp?qkn+sUt\l54 Q tzq)w)sUoMHUj y\sUtJu o\tJj \j<oJw)n9lWqkn( oJj y\jn ≈ h−1

2

xsUo\wpflGl)o\wGj<wo\t\ju<sUt HUj<wG*Uj<tJu<jy\jz~9p 85qkJs y\jy\jx<tFp?~9n9l)p?qkn+sUt j5t y\n+j5t\l)n+sUt14

b-.IpfoJq)w)j<lZqkj5lGqklBj<ty\n+j5t\l)n+sUtxJ~9o\lZ<~+j!HU5j lGj<w)pfn9j5tqBt\<u5j<lGlkpfn9wGj<lMxsUo\w0HUsUn+wl)nu5j5q)q)j \j<o\wGn9lWqkno\jy\5xj<t\yy\j ~_p y\n98j<t\lGn9sUt:4

& !##

bj ~_p 87<8j pft\n(<wGjo\j xsUo\wB~_p 85qkJs y\j yJj x 5tFpf~+n9l)p?qkn+sUt#;JsUtz= pfn q H?pfwGn9j5wB~9j u5s j 8u5n9j<tq y\jw)5*Uo\~9pfw)n+lkp?q)n+sUt

nj<t87<8jZq)j<8x\l oJj

N, δjqM

yFpftJl ~9jZu|pfl y\jB~_p p?l)j . $ -%<3?: 8P4 Q txJw)j<tJyqksfoGsUo\wGl

n = n0(√

2)(j−1)αnxsfo\w

n0 = 0.5j5qZsUtqkj5lGqkj>~9p-u5sUt-HUj5w)*Uj5t\u<j y\jm~/.Ipf~9*fsUwGn q)\8j xsUoJw

y\n :5w)j<tqkj5l=H?p?~9j5o\w)lBy\jαn4

j

Prix

105 15

14

15

16

17

13.5

14.5

15.5

16.5

17.5 0.1 0.5 1 2 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

αn " + . # " ; = %

". "αn

;2: .") " 4 .! *( ") " = ". :!3 " "8) ." .) ; , =): ."; . "; 6 " "

n 5 =( " " " &" '(

&.; .) ; . 1

% .; . %; n

. $ ); " 2 " "!: " " ( = ) ." .) ; &-

Y =@

" #( ; " ; ) ." $n

: )" n

; )?." "3 2 = " " = @ /1. % %&." ! "

n ; ! " )"" =)5 " ) "

" 2 " / " =φn

" ) ." ) " % " ;

+

Page 192: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

e c' BdMa h 4 Q Z' Q &mV%cdM'We ' &a V

)=%: " ?)*!" 0 $ ')') $&%? " "#/1 >Q txj5oJqc<*pf~+j<8j<tq o3qkn9~+n+l)j5w ~/.Ipf~+*Usfw)n q)\8j xsUo\w y\5q)j<wGn+t\j<w oJt qkj<8x\l y#.Ep?w)w)7qcsUx3qkn9pf~ lGo\nJHrpftq~_p 85qkJs y\j y\j `sUt\*flGqCpT V3u 0pfwWq>X?4 d pfxJxj<~9sft\lB~+j>x\w)n+t\u<n+xj>y\j u<jq)qkj qk\s yJjQ tn9tJn+qkn9p?~+n l)j>p HUj<u

yN,MN (·) = g(tN , ·)jqMxsUo\wZq)sUoJq

msftxsUl)j

τm = tN4

Q t u<pf~9u5o\~9j xFpfwoJt\jw)<*fw)j<lGl)n+sUt ~9n+t\<pfn+w)j

yN,Mk (·) x\o\n9lxsUoJwq)sUoJq m ;lGn yN,Mk (XN,mtk

) ≤g(tk, X

N,mtk

)pf~+sUw)l6sUtzpfu5q)oFpf~+n9l)j~+jqkj5xJlZy#.Ipfw)wG75qZsUxJqkn+pf~xsUo\w~_p(qkw)pGj5u5q)sUn9wGj

mj5tzx sflkpftq

τm = tk4

tJj =sUn+lZxFp?w HUj5t o ,~ .In+t\lGqkpftq

t0;FsUtzu|pf~+u<o\~+j>~9j>xJw)n 2pf8<w)n+u|pfn+tu<sU8j

1

M

M∑

m=1

exp(−rτm)g(τm, XN,mτm ).

& sfo\lZxsUoMHUsUt\lBpfx\x\~+no\j5wu<jq pf~+*UsUwGn qk\jj5ty\qkj<wGn+tFpftqB, u \po\j n9t\lWqCpftqyN,Mk (·) xFpfwo\t\jw)5*Uw)j5l)lGn9sUtl)oJwM~9p pfl)j. $4csUn+u<n#~+j<lZw)<lGo\~+qkp?qkl

j

Prix

10512

13

14

15

16

17

18

19

1 1.5 2 2.5 3 Prix de référence

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

j

Couver

ture

105

-0.6

-0.5

-0.4

-0.55

-0.45

1 1.52 2.53

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1%& b?c0+*$M<>q +, ^7|BVMj*._&O*,=&0 ^.B,OM=&*,&).

5(&L0&*,=&,0..=9-*(=g1*00* + B

αMX?]&1,Q+*.1**.112*,9

αδ,&*,(J1*/<CJ)(&*+&

& 1*.0&0*j17*R'&(&*.* 1*\4B&<>10Q17*αM = 1

1.5

VN+ ?A+*,$.M<>(*+L0&(&*,gJ1*,$4Q*=\=&*,&F17*

αM,&L+*.XN

Page 193: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

` .Ep?~9*UsUwGn qkJ8jvy\sft\t\jo\t\ju<sfoMHUj<wWqko\w)j X xsUo\w

αMp?l)l)j Xz*Uwkp?t\y ^(y\o 75jvsUw)yJw)j o\ju<j5~9~9j

s q)j<to\j xFp?wB~9j5lMp?oJqkwGj<lBqk\s y\j5l<4

i1D

Page 194: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

$ % B

$ ' *!) " $ /;!) *B24/ *!) "#/ @')(* , 0*!) /;

Q tH?p q)wkpfn qkj5w~9jmu|p?lBy\o8s y5~9j= pfuqksUwGn9j5~:xsUo\wZ~ .I HUsU~9o3qkn9sfty\j<lZ=sUwW0pfwGy\lZl)o\wB5~9j5u5q)w)n9u5n+q)mxsUo\w8sUtq)w)j<w o\j5~9l8l)sftq~+j<lwG<l)oJ~+qCprqkl oJj~9pqk\s y\jzy\sft\t\jyFpftJl8u<ju|p?l<4 a t xFpfwWqkn9u5o\~+n9j5w<; sUty\5l)n9wGj Hrpf~9sfw)n9lGj<wo\t\j sUxJq)n9sUtu|pf~+~#j<o\wGsUx<j5t y\j p?q)o\w)n qk

T = 10GsUoJw)l<4 `j l)sUoJlWT()p?u<j<tq j<lWq o\t

u<sftqkw)p?qZ=sUwG06pfw)yzy\j y3@3t\pf8no\j

dF (t, T )

F (t, T )= σe−a(T−t)dWt.

`j<l Hrpf~9j5o\w)ly\j<lvx\pfwkpf 5q)w)j5ll)sUtqa = 0.19

;σ = 0.23

jq~+jqkpfo32 y#.Ln9tqk5w)75qr = 0.05

j<tH?p?~9j<oJwp?t\t oJj<~9~+jf4 ` p H?pf~9j5o\wMy\o lWqkw)n+Sfj(j5lGq

K = 100jq~9p Hrpf~9j5o\wMn9tJn+qkn9p?~+j>y\o lGsUo\lWT )pfu5j<tq H?p?oJq

F (0, T ) = 1004

pf~+sUw)n+l)j5w(~ .LsUxJqkn+sUt w)j H n9j5tq ,qkwGsUoMHUj<w-~9ju5sUo\x\~+jvy\jxJw)s u<j5l)lGo\l(Y, Z)

lkp?q)n9lG= p?n9lkp?tq(~/.L oFprqkn9sftw)qkwGsU*Uwkp?y\j

Yt = (F (T, T ) −K)+ +

∫ T

t

f(s, Ys, Zs)ds−∫ T

t

ZsdWs

sUf(t, y, z) = −(y − z

σea(T−t))r

4 Q t p?x\x\~+no\j~ .Ep?~9*Usfw)n qkJ8j8yJ<u5w)n+q-yFp?t\l-~_pxFpfwWqkn9j8'G' p HUj<u~9j5l pfl)j5l . $ j5q . $ -%<3?:48 jq t\sfo\l-w)j<xsfwGqksft\l-~9j5l-wG<lGo\~+qkp?qkl(s q)j<t oJl xsUo\w(o\t\jvu5sUoMHUj5wGqko\wGjGsUo\wGtFpf~+n5w)j

(N = 10)jqZo\tJjmqkpfn9~+~ j y:.I@3xj5w)u5o j<l

δ = 2x\o\n9l6xsUo\w6o\tJj>u5sUoMHUj5wGqkoJw)j D =sUn+l6x\pfw

GsUo\w(N = 20)

j5qMo\tJj qkpfn9~+~+jy#.I@ xj<w)u5o j<lδ = 1

4

Page 195: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

' 4cd a V `MV& acdB' a V

M

Prix

41. 10 43. 10 45. 10

11

12

13

14

15

11.5

12.5

13.5

14.5

N=10,delta=2,HC(1,0)N=10,delta=2,HC N=20,delta=1,HC N=20,delta=1,HC(1,0)

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1

M

Std Pr

ix

41. 10 43. 10 45. 10

0.1

0.2

0.3

0.4

0.5

N=10,delta=2,HC(1,0)N=10,delta=2,HC N=20,delta=1,HC N=20,delta=1,HC(1,0)

P0

t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 1*,$4;>*=\=&*&C17*LC1XWX& 1*LYSW &%

14.72N *,(C*

j&*7JU_&g<>=&0*=0*,H?cB.00R%HC(1,0)

Jb?]=&&*,iGH17&J1U<>17* )12*

(N, δ) = (20, 1)'17*

(10, 2)P&&*;5(-0&U*=$

R'1*.0J, N

!"$#%$&('*)+,$.-!/0&-1243 65748-)+:9-

;=9".&.= :V4B&<>102=&*,& B46X?c0=&*\>KR"1*,.01*=&=&&F%J&=&**%9(& &C&F0&@N;&17&)8WF M+&%$.0.*<>0

Φ(S) = (ST −K1)+ − 2(ST −K2)

+ N QVM$4Q1*.<>/*&&

µ σ r R T S0 K1 K2< N <>= < N*) < N < < N < < N*) = ?<@< A= B<A=

C=&,+g1*BSi(r)

1*,$4> B M0&q=(=8(&0,.*,Ki

_.B4KX?c0=&*\rN

BT.LQ1*,=&0*Y0

1*L&<',0=80*L1*$4_R B M&W

'

Page 196: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

BS1(R) − 2BS2(R) "!BS1(r) − 2BS2(r) $#BS1(r) − 2BS2(R) % '& BS1(R) − 2BS2(r) ( '#")

*+-,/.102.1354$*6,+-*7,/8:9<;"=5><?A@CB:@EDF;"?AGH4I=/4$.135;"*KJML' NOPQ&RTSVUXWY;"*7GZ@<+-,["9F=\.1]H@<=^6+:@C02@<3,/=\;".23G:=\@X_?A.2>F=\@F3O01.2`"*-@F3O3\;"*R,aB-@F3CbZ;"=\*:@F3E.2*-cd9<=5.1@F+:=\@F3OGZ;"+:=

Y0

4$01;"=53Q^e+-@f0V4gB:@F=\*:.2><[email protected]`"*:@acd;"+:=5*-.1,O+:*:@bZ;"=\*:@O35+:GZ9<=5.2@<+:=5@Ih*i=5@<GZ;"=\,/@jD<.k_lB:@F3\35;"+:3Q024m?A;on"@<*:*:@jB:@F3QG:=\.kp;$b-,/@<*e+:3QGH4$=0rq'4$01`";"=\.k,58:?A@C3\+:=

50024$D<@F?A@F*6,53@X,

@F*R,/=\@G:4$=5@<*6,/8->F35@<3s0rqt9FDF4$=\,M_l,nGZ@u@F?AG:.1=5.1^e+-@$WGZ;"+:=fB-.1vw9F=5@<*6,53A,neGZ@<3sB-@ucd;"*:D<,5.2;"*-3AB:@7bH4$3\@@X,B:.kvx9<=5@F*R,/@<3[I4I02@<+:=53B:@

M

N = 5 N = 20 N = 50

M D = [60, 140] D = [60, 200] D = [40, 200]

δ = 5 δ = 1 δ = 0.5

%y$z ( )R!-J)- "U ( % J) &"!"U ( '#"&HJd|: % !"U! %o & ( J)- %"% U ( % |YJ) % #RU ( t| z J)- !I|6U

)$| z & J)-')R!"U ( ')")HJ) ) ( U ( ') z J)- %o Uz-% & & % J)-') ( U '&"#HJ) ) U '&"&HJ)-') U

(R I# z &"):J)-') % U '&R!:J) ) % U '&"#HJ)-') % U

~ e % 9F3\+:01,/4I,/3;"b-,5@F*e+-34y["@<DE0V4jbH4$3\@Cif

bH4$3\@O b:4$35@Q bH4$35@Q bH4$3\@OmFHkM 16

=59<`".2;$*:3B:@Q;"=5;$*:;".64

=5@F`-:;"=<10

=5@<`::;$=F10

=5@<`:-;"=FN = 5 N = 20 N = 20 N = 20

%o$z ( I( J)- ( )RU |: !$)HJ % % U ( ') z J)- !"U ( $( J)- $( U! %o ( )"!:J)- %y( U ( ( #HJ)- % )RU '& % J)- %"% U ( ) ( J)-') z U

)$| z &I|YJ)-')"#RU ( ')R!:J)- )I|6U '&"):J)-')"#RU &R-J)-')$|6Uz-% & & J)-') ( U '&"#HJ)- ) U z #:J)-') ( U &R!-J)-') U

(R $# z &$)HJ)-') U '&$|YJ)- ) % U z #:J)-') U &R!-J)-') % U

~ E 9<35+:0k,4I,53;"b-,5@F*6+:34y[$@FDE01@F3b:4$35@<3@X,sFHkxo

+02@F3Q=\9F3\+:01,/4I,/3QB:@Ai@<,QfWYN,M

t0

35@F?jb:01@D<;"*6["@<=5`"@<=O["@<=532.95

@<,5,5@[T4$02@<+:=Q*:@GZ@F+-,GH4$3<,5=5@G-=59FB-.1,/@GH4I=+:*:@D<;"?jb:.2*H4I.23\;"*A02.2*-94$.1=\@B-@G:=5.pf0V4ID "_MeD 8:;$02@<3¡eBH4$*:3DF@<,@Xp@F?AG:02@IWe0V4*:;"*e_02.1*:9F4$=5.,/9QB:@

f4A+:*=59<@F0.1?fG:4$D<,35+:=

Y0

Zq'.1*6,/@<=5G:=\9<,/4I,/.1;"*]H*H4I*:DF.1>F=5@E@F3\,^6+:@C02@Q["@F*:B-@F+:=B:@a0rqt;"G-,5.2;"*7B:;$.1,4I01,/@<=5*H4T,/.k["@F?A@F*6,@<?fG:=\+:*6,5@F=@<,G:=\<,/@<=B:@a0rq'4$=\`"@F*R,GZ;"+:==\9FG:01.2^6+:@F=02@CGH4n";$vB:@O0rqt;"G-,5.2;$*

% &R!

Page 197: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

lj 7 MP

N = 5 N = 20 N = 50 N = 50

M dy = 1Wdz = 0 dy = 2

Wdz = 1 dy = 4

Wdz = 2 dy = 9

Wdz = 9

%o$z z :J)- ( &RU ( ) % J)- |6U ( ') J)- " U ( t|R&HJ % !""U! %o zR J)- )RU &$|HJ)- %o U '&R-J)-')"&RU ( ') J) % U

)I| z z J)-')R"U & J)-')R"U '& J)-')$|6U '&R:J)-') ( Uz-% & z J)-')R!"U & J)-')$|6U '& J)-') U '&"#HJ)-') % U

(" $# z $&HJ)-') ( U & % J)-') U '& % J)-') % U '&R!:J)-') % U

~ (C 9F35+-01,4T,/3;$b-,/@<*e+:34["@FDO0V4jbH4$35@gA

*gD<;"?AGH4$=/4I*6,02@F3B:.kvx9<=5@F*R,/3D8:;".pfB:@cd;"*:DX,/.2;$*:3B:@bH4$35@IW;"*sGZ@F+,=\@F?f4$=5^6+:@<=^6+:@024CD<;"02;$*:*:@N = 5

B:@C Jd4$b-02@F4$+ UG:=\9F3\@F*6,5@B:@<3=59<35+:0k,4I,5335.1?f.1024I.1=\@F34y["@<Dj+:* *:;$?b-=5@C9F^6+:.k[I4$01@F*6,B:@cd;"*:DX,/.1;"*:3OB:@AbH4$3\@f^6+:@f024D<;"02;$*:*:@

N = 5B:@f Jd4$b:02@F4$+ e % UX4$*:3E02@4$b:01@4I+ W

02@<3B:@F+epmB:@F=\*:.2><=5@F3D<;"02;$*:*:@F3?A;"*6,5=5@<*6,^6+qt+-,5.202.135@<=+:*:@b:4$35@B:@GZ;"01ne*$?f@<302;eDF4$+pAGZ@F+,4$?A9X_02.1;"=\@F=C35.2`$*:.1]:D4I,5.1["@<?f@<*6,j024 G-=59FD<.23\.2;"*x ;$+:34y[e.2;$*:3jB:9 ;"b:35@<=\["9D<@F0V4 B:4$*:30V43\;"+:3M_l35@<D<,/.1;"*|: GZ;"+:=02@<3bH4$3\@F3i @<,sFHky .1*H4$01@F?A@<*6,W6[e+02@4Ib:02@F4$+u ( W-024abH4$3\@OB-@QGZ;"0kn*"?f@<3`"01;"bH4$+epGH4$=\[e.1@F*R,a9<`R4$01@F?A@F*R,74T,5,/@<.2*:B-=5@A024g[I4$01@F+:=C4I,\,/@<*:B:+:@IW02;"=\35^6+:@A*:;"+:3a4$+-`"?A@F*6,5;"*:3E02@*:;"?jb-=5@fB:@mGZ;$01ne*"?A@F3EBH4$*:3a024ubH4$35@I\;"+,/;"*:3C^6+:@sGZ;$+:=C0V4uB:@<=5*:.1>F=5@fDF;"01;"*:*:@AB:+ 4$b:01@4$+ ( W02@<3gb:4$35@<3 0q'.1*:3\,/4$*6,

tk*:@i3\;"*6,G:4$3 J4["@FD

d = q = 1Up0,k = 1, x, · · · , xdy

Jd=5@F3\Gp1,k = 1, · · · , xdz

Ug?f4$.131,

x−minm SN,mtk

maxm SN,mtk

−minm SN,mtk

, · · · , ( x−minm SN,mtk

maxm SN,mtk

−minm SN,mtk

)dyJ@X,70V4 ?f<?f@

D8:;"35@jGZ;"+:=p1,k

UX@X,5,/@j*:;"=\?m4I02.1354I,/.k;"*7@F3M,Q*:9FD<@F353/4I.2=5@j02;"=\35^6+:@a02@a*-;"?jb:=5@aB-@jGZ;"01ne*"?A@<3B:@X_[e.2@<*6,`$=/4$*:BxWHGZ;$+:=9<[e.k,/@F=Bq 4[";".1=B:@F3?m4I,5=5.1DF@F3B:@O=59<`"=5@<3535.1;"*u?m4I0D<;"*:B:.k,/.1;"*:*:9<@F3F

! "#$ &%('*)+

h* DF;$*:35.1B:>F=\@,5;"+,\;"+:=\3d = q = 1

@X,A02@g?A;eB:><02@sB:@u0V4$D"_MD8:;$02@F3GZ;"+-=S

4iDF;"*:B-.1,/.1;"*,/@<=5?A.2*H4I01@B:@A0q Q [I4$+-,uG:=\9F3\@F*6,

Φ(S) = ( 1T

∫ T

0Stdt − K)+

W 35;$.1,O02@DF4$3EBqt+:*:@A;"G-,/.1;"*4$3\.V4I,5.2^6+:@$W-@<,*-;"+:3G:=5@<*:;"*:301@F3GH4$=54$?f>X,/=\@F33\+:.1[T4$*6,53E¡

µ σ r T S0 K)-')"# )- )- % % % )") % )");"+:=4$G:G:=\;eD 8-@F=DF@X,5,5@ED<;"*:B:.k,/.2;$*s,/@<=5?A.2*H4$01@GH4I,58_B:@<GZ@F*:B-@F*6,FWYDF;"?A?A@Q3\+:`"`"9<=59OBH4$*-3024jGH4$=\,5.2@WH*:;"+:3G:=\@F*:;"*-3

d′ = 2@<,35.2?j+:01;"*:3

PNtk

= (SNtk ,1

k+1

∑ki=0 S

Nti

)∗Jd[";".1=ELte) % SVUXY@<3=\9F3\+:01,/4I,/3

G:=\9F35@<*6,59F3QBH4$*:302@E,4Ib:02@F4$+it|s35;$*6,DF;"8-9F=5@<*6,53QD4I=02@CG:=\.kp7B:;$*:*:9jG:4$=0q 4$01`";$=5.,/8:?@E*qt@F3\,QGH4I3,/=\>F39F01;".2`$*:9B:+uG:=5.pgB:@O=59Xcd9F=5@<*:DF@

7.04B:;"*:*-9EBH4I*:3EL'e) % S

;"?A?A@?A@F*R,/.1;"*:*:9BH4$*:3Lt) % SrWy0rqt+-,5.201.k3/4I,5.1;"*QB:@ 1N+1

∑Ni=0 S

Nti

GZ;"+:=4$G:G-=5;eD8:@F= 1T

∫ T

0Stdt

@<3\,02;$.2*CBqt<,5=5@;"G-,5.2?f4$0ThQ*jGZ@F+, ["9F=\.1]H@<= D<@^6+:.35@GH4$3\35@3\.-;$*D8H4$*-`"@

PNGZ;"+:=?A.2@<+pa4$G:G:=\;eD 8-@F=

1T

∫ T

0Stdt

;$?f?A@ G:=5;"GZ;$359 BH4$*-3 L'e) % SW *:;"+:3 4$G-G:=5;eD8:;"*:3 1T

∫ T

0Stdt

G:4$=

% &"#

Page 198: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

N = 5 N = 20 N = 50

M δ = 5 δ = 1 δ = 0.5

D = [60, 200]2 D = [60, 200]2 D = [60, 200]2

%y$z #- ("( J)'| % U |:t|6:J ( z "U ( t| z J %y( ') z U! %o #-'#R!:J) % U #- $z J)I#RU !'# ( J ( "U

)$| z #- z )HJ) )"&"U #- $#HJ) |6U #-t| z J)-'|"&RUz-% & #- z"( J) )$|RU #-'&R!:J) )$#RU #- z #HJ)- %y U

(R I# z #- z"( J) ) U #-'& z J) ) ( U #-'&"&HJ)- )I|6U

~ '| 9<35+:0k,4I,/3GZ;"+:=0rqt;"G-,5.2;"*g4$35.24I,/.1^e+-@Q@F*u+-,/.102.1354$*R,0V4jbH4$3\@Cim

1N

∑N−1i=0 SNti (1 + µh

2+ σ

2∆Wti)

WFDF@^6+:."DF;"*:B-+:.1,PNtk

=(SNtk ,

1k

∑k−1i=0 S

Nti

(1 + µh2

+ σ2∆Wti)

)∗GZ;"+:=

k ≥ 16 @F3=59<35+:0k,4I,53Jd[";$.2= ,4$b:01@4I+m !"U 35;"*R,b:.1@F*A?f@<.20101@<+:=534["@FDD<@D 8:;$.kpGZ;"+:=

PN @

G:01+:3FWy*:;"+:3;"b:3\@F=M[";"*:30V4DF;"8:9<=5@<*:DF@B-@0q 4I02`";$=5.,/8-?A@^6+:.6G:=5@<*:Ba@<*a.1*:G:+-,B:@<335.1?j+:0V4T,/.k;"*:3 B:@SN35;$+:3024CG:=\;"bH4$b-.202.,/98-.23\,5;"=5.1^6+:@jJ

µ 6= rU@X,DF;"=\=5.2`$@02@B-=5.1c ,GZ;"+-=B:;"*:*-@F=02@G:=\.kpf=\.23\^e+-@X_*-@F+-,5=5@$

z (" %o$z ! %o )$| z z-% & (R $# z=5.p # )-'&") |:t|R& #-'# z # % ! #- z"z # &"& e ) DF4$=\,M_l,nGZ@ |- ) z z ) %"% |:'#$| % %$% )- % ) )R ) ) ~ e! 9F35+-01,4T,/3GZ;$+:=0rqt;"G-,5.2;$*4$35.24I,/.1^6+:@Q@F*u+-,/.102.k3/4$*R,+:*:@O?f@<.20101@<+:=5@Q4$G:G-=5;yp.1?m4T,/.1;"*sB:@1T

∫ T

0Stdt

@<,024b:4$35@E JN = 20

Wδ = 1

WD = [60, 200]2

U

! $*)+!) $ '*)+ )*) * )+#

4G:02+-GH4$=\,B:@<3@Xp@<?fG-02@F3^6+:@*:;"+:34y[";"*:3,5=/4$.k,/9<3\+:3\^e+xq G:=59<35@F*R, 35;"*R,@Xvx@<D<,/+-9F3@F*aB:.1?f@<*:35.1;"*1

R;"+:34$0102;$*:3.10201+-3\,/=\@F=3\+:=+:*@ p-@<?fG-02@02@<3G-=5;"b:01>F?A@F3 B:@0V4B:.2?A@<*:35.1;"*";$?f?A@bZ@<*:D8:?f4$=5YW*:;"+-3D8:;".135.13535;$*:3 m*:;$+-["@4$+ +:*i@ p@F?AG:02@E,/.2=\9CB:@sL')$|$S?f4$.13DF@<,\,/@Ccd;".23_DF.@F* B:.1?f@<*:35.1;"* ( S

35+-.1,0V4jB-ne*H4$?A.2^6+:@O0V4$D"_MeD 8:;$02@<3@<*7B:.1?f@<*:35.1;"* ( @X,02@QGH4n";$v [I4$+-,(K − (

∏3i=1 S

it)

13 )+

@F3[I4$02@<+:=53B-@F3GH4$=54$?A><,/=\@F33\;"*R,

µ = 0.05Wr = 0.05

WT = 1

WY02@<3DF;$?fGZ;"354$*R,/@F3B:@S0

[I4I02@<*6,

100@X,K = 100

H4?m4T,/=5.1DF@σ

[I4$+,σ =

0.4 0 0

0 0.4 0

0 0 0.4

H @QG:=5.pgB-@O=59Xcd9F=\@F*:D<@EB:@QDF@X,5,/@

;"G-,5.2;"*7B:;$*:*:9aBH4$*-3jL')$|$S@F3\, w;"+:3B:9<35.2=\;"*:39X,4$b-02.2=B:@<3bZ;$=5*:@<3.1*-cd9F=\.2@<+:=5@<3GZ;"+:=DF@G:=\.kpg4$?A9F=\.2DF4$.1*s@X,+-,/.102.13\;"*:3B:;"*:DQ0q 4$01`";$=5.,/8:?@4y["@<DOGZ9<*H4$02.13/4I,5.2;$*s^6+:@O*-;"+:3D 8-;".23\.2353\;"*:39<`R4$02@A;$+:3AG-=59F3\@F*R,/;"*:3A,5;"+-,Bq 4IbZ;"=5B B:@<3A=59F3\+:01,/4I,/34y["@FD0V4ibH4I35@7i @F* D 8:;$.23\.235354$*6,

αδ = 1@X,

αM = 4.5 * @<vw@<,0247G:=\;"GZ;"35.k,/.1;"* lj % .1?AGw;$35@

MB-@m[I4$=\.2@<=aD<;"?A?f@

h−(1+β+αδd′)

4["@FD

% &R

Page 199: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

lj 7 MP

β > 1DF.D<@F0V4 B:;$*:*:@i^6+:@

MB:;".k,s,5@F*:B:=\@i["@<=530q'.1*-]H*:.G-02+:3m[e.1,/@ ^6+:@

h−5;"+:34[";"*:3

9F`"4$02@<?f@<*6,D<;"*:3\,/4I,/9^6+:@D<@35@F+-.209<,/4$.1,,5=5;"GjGZ@<3535.1?f.13\,5@GH4$==54$G:GZ;"=M,4$+epj=\9F35+-01,4T,/3*e+-?f9<=5.2^6+:@<3;"b:3\@F=M["9F3Feqt@F3M,024=/4I.235;$*jGw;$+:= 0V4I^e+-@F0201@*:;"+:3 G:=5@<*:;"*:3

αM = 4.5";"+:3 ;"b-,5@F*:;"*-302@<3 =59<35+:0k,4I,53

35+-.1[I4I*6,/3O¡

j

Prix

1 2 3 4 5 6 7

9

10

11

9.5

10.5

Méthode de pénalisationPrix théorique

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1 !"$#%&')(+*,&- .0/1 ')233!

4'5*&'6(52,!'"7&8j = 7

!9:/1*!; '"<9=!9*9&')(5'>?,!'" &'6(2,'@,AB CD*,@E:/1F'"'"@3,!2;*@*@&8,!- "' G'"IHJK'"':L;K&')(*,&- 0/1 ')235MEN"'"*I/O P'"K!!,.HRQ/1!S:/1ATUA@*V *'3'P 'W+')X"3&= !Y33Z&= 3YZ*'3'

1![& !3Y!"

. .'V .=\*U/1,AB .(∏3

i=1 Si)

13N='"C9!9," .<'). C9!Y]'"'" C%*I/1D&- .9!9= !

^`_a!^b_c!dfe)gUhij*'3'1 k!5A5!"@. 'V ="@*I/l,!B G5#

j

Prix

105

9

10

11

9.5

10.5

11.5HC HC(1,0) Prix théorique

P 0 t 0 t 1 t 2 t 3 t 4 t 5PN,

1t 1 PN,

1t 1 PN,

2t 1 PN,

2t 1 PN,

3t 1 PN,

3t 1mRnSopqrstusqosKs!vw\xy"sq3z3sy")s|tsDz3z3s~vyf"s~ty)NqI1n5-n.vWsqoptsiwp|w\Eny)wo!pqrs!tui8p|t

s!vstqy!tsv9r.n"s|tvsjws!vw!svpqNpxvs!trs~sqopts~s!vty6"u!ts!zs!qwv|8!ty"s|tvWn|Yty)

wpty|skK y"qvy:-s5tpx!zsSsX01n|uz3sqCwBn.wy"pqsXVn3yz3sqvypqjs!vw@sXwtp|]rstsXx8pqqs!vr.ntynx"svIl!wn.wzFn.tpr;y"sqqsv9p|Ns4x8pqqs!v?pqo!wypqv<s4x-nvsD8p|t9Ar;y"ws!t9I1nkrpy"t9@|wy"y)vst%s!vxnvs!vsD?pqoAwypqvt|y"zs!qwnytsv<ws!s!v`|y-os!twsvWopqCrstusqCw7zny"vE|qo!pwWs~oPn.o|-!s!r

k

Page 200: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

Page 201: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe
Page 202: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

;$+:34y[";$*:39<,5+:B:.29BH4$*-3DF@<,\,/@,/8:><35@+-*:@?f9X,/8:;eB-@B:@u;"*6,5@X_M4I=502;EGZ@<=5?A@<,5,/4$*6,Bq'4$G:G-=5;eD8:@F=0V4 35;"01+-,/.1;"* Bq'+-*:@ 9F^6+H4I,5.2;"* B-.1vw9F=5@<*6,5.2@F0101@ 3\,5;eD 8H4I3\,/.1^6+:@ =59<,5=5;"`$=/4$B:@i?m4I=5$;o[e.1@F*:*:@I+4$*-3u0V4GH4$=M,/.1@ W*-;"+:34[";"*:34I*H4$01ne3\9 01@F3u@<=5=5@<+:=537B:@<3 4I+ G:4$37B:@,/@<?fG:3

hW 0rqt+-,5.201.k3/4I,5.1;"*Bqt+:*

*:;"?jb-=5@]H*:.HB:@cd;"*-D<,/.1;"*:3B:@bH4$3\@K

@X, O+:*A*:;$?b-=5@]H*:.M

B-@3\.2?j+:024I,5.1;$*:3 B:@7;"*R,/@X_4$=501;:;"+-,5@<cd;".13FWo02@<3bw;$=5*:@<3GZ;"+-=0q'@<=5=5@<+:=B:@

M*qt9<,/4$*6,GH4$34$3\35@@XpG:02.1DF.k,/@<3GZ;"+:=+:*:@+-,/.102.1354I,/.k;"*

@F*G-=/4I,5.2^6+:@$WH*:;"+:34[";"*:34I?f9<02.1;"=\9QDF@F3@F3M,/.1?m4I,5.2;$*:3GZ;$+:=+:* 4$01`";"=\.k,/8:?@O019F`"><=5@<?f@<*6,?A;eB:.k]H9BH4$*-3024G:4$=\,5.2@\6@<3*:;"+-["@<0202@<3bZ;$=5*:@<3*:;"+-3;"*R,4I02;"=\3 GZ@<=5?A.23 [email protected]=5@[I4I=5.1@F= 35.1?j+:01,/4$*:9<?A@F*6,h

W"01@*:;"?jb:=\@B:@cd;"*:D<,5.2;"*-3 B:@bH4$35@@<,M

35.1?j+:01,/4$*:9<?f@<*6,F$;$+:34y[";"*-3["9<=5.1]:9*e+:?A9F=\.2^6+:@<?f@<*6,0V4[I4I02.1B:.k,/9QB:@CDF@<3@F3\,5.2?f4I,5.2;$*:3BH4$*-3024GH4$=\,5.2@Oljx@<GZ@F*-BH4$*6,FWwDF@<,\,/@CB:@F=\*:.2><=5@CGH4I=\,/.1@E3\;"+:01><["@B:@O*:;"?jb:=\@F+:3\@F3^6+:@<3\,/.1;"*:3O¡

;$+:3E4y[";"*:3O9X,4$b:01. B:@<3E=59<35+:0k,4I,53OB:@bZ;"=\*:@F3OGw;$+:=O024g*-;"=5?A@L2

B:@0rqt@F=5=\@F+:=QDF@^6+:.*:;"+:3E4DF;$*:B:+:.k, E024$*:D<@F=<W6GZ;"+:=D 8H4I^e+-@\@<+mB:@GH4$=54$?A><,/=\@F3

(N,M,K)W!$)Qcd;".130rq'4$02`$;"=\.k,58:?A@B-@?f4T_

*:.1>F=5@+["9<=5.k]H@F=0V4[T4$02.1B:.k,/9B:@D<@F3bZ;"=5*:@<3F6 *AG:=/4T,/.2^6+:@IW";"**:@c4I.1, GH4$3,5;"+:=\*:@F=0rq'4$02`$;"=\.k,58:?A@!$)ucd;".23EGZ;"+:=a;"b-,5@F*:.1=C+:* G:=5.p @<,j+:*:@mDF;"+-["@<=\,5+:=5@I .2*:3\.rW.20D<;"*6[e.1@F*:B-=/4$.k,aBq'9X,4$b-02.2=E9<`R4$02@ _?A@F*6,QB:@<3O=59<35+:0k,4I,53QB:@DF;$*6["@<=5`"@<*:DF@AGt3FGZ;"+:=0rq'4$02`$;"=\.k,58:?A@IWw^6+:. GZ;"+:=5=54$.1,9X["@F*R,/+:@<0201@F?A@F*6,G:=5;y["@F*-.2=B:@F3bZ;"=\*:@F3GZ;"+:=024Q*-;"=5?A@

L202;"=\35^6+:@DF@<3bZ;"=5*-@F3,5@F*:B:@<*6,3\+A3/4$?A?A@F*R,[.k,/@["@F=\3

0

;$+:3E4y[";$*:3Q,/@F3M,/9DF@<=\,4I.2*:3QD 8-;".kp B:@bH4$3\@F3EB:@jcd;"*:D<,5.2;"*-3Q^6+:.*:;"+-3OGH4$=/4I.235354$.2@<*6,QD<;"8:9F=\@F*R,/3?f4$.23 B:@*:;"?jb-=5@F+epf4I+-,/=\@F335;$*6, Q,5@F3M,/@F=<e *fGH4I=\,/.1DF+:01.1@F=<W$+:*fD8:;".kp*H4I,5+:=5@<0Y35@<=/4I.1,B:@B:9<]H*-.2=02@O35+-G:GZ;"=\,Bxq'+:*-@ab:4$35@CBqt86neGZ@<=5DF+-bw@F3FWwB:@C?f4$*:.1>F=\@"ADF@C^6+:@CD8H4$^6+:@a8RneGZ@F=5D<+:bZ@jB-@C0V4b:4$35@DF;$*6,/.1@F*:*-@i01@?f<?A@*:;"?jb:=\@ B:@ 35.1?j+:0V4T,/.k;"*:3 K+:* .2*:3M,4$*R,gB-;"*:*:9I0c4I+-,4I02;"=\3m9X,/+:B:.1@F=0rqt@F=\=5@<+:=Bxq 4$G-G:=5;p-.1?m4T,/.1;"*sBq'+-*:@Ocd;"*:DX,/.1;"*u .2G-35D8:.1,OG:4$=+-*:@O,5@F0102@bH4$3\@$

q 4I02`";"=\.k,/8-?A@*:9FD<@F3\35.1,5@^6+:@0V4ODF;"*:B-.1,/.1;"*,/@F=\?f.1*H4$01@B:@0rq' Q-3\;".1,+-*:@cd;"*:DX,/.1;"*fB:9<,5@F=5?A._*:.13\,/@QB:@O024a[I4$01@F+:=]H*H4I02@QBqt+:*:@Q[I4I=5.V4Ib:02@Bxq'9X,4I,?m4I=5$;o[e.1@F*:*-@$Y@FGZ@<*:BH4$*R,WYDF@X,5,5@ED<;"*:B:.k,/.1;"*@F3M,G:4$=\cd;".13DF;-,5@F+:3\@+D4I+:35@B:@024B-.2?A@F*:3\.2;"*CB:@024[T4$=5.24$b:01@Bqt9<,4T, ?m4I=5$;o[e.1@F*:*:@@X, .20635@<=/4$.k,.2*R,/9<=5@F3\3/4$*R,OB-@j,5@F3\,5@F=QB-@F3O4$0k,/@<=5*H4I,5.1["@<3DF;"?A?A@aB:@jG:=5@<*:B:=\@GZ;"+:=01@F3=\9F`"=\@F3\35.2;$*:3QB:@<3cd;"*:D _,/.1;"*:3B:@b:4$35@^6+:.:3\;"*6, cd;"*:DX,/.1;"*:3 Bxq'+:*-@4I+-,/=\@[T4$=5.24$b:01@Bqt9<,/4I,W$GH4$3cd;"=5D<9F?A@F*R,?m4I=5$;o[e.1@F*:*-@$W@<,[email protected]?A@F*:3\.2;"*gG:01+:3GZ@X,/.k,/@$

;$+:34[";"*:3s,5@F3\,59iG:02+-35.2@<+:=53s?A9<,58:;eB:@<3gBq'4$G:G:=\;yp.2?f4I,5.2;$* GZ;"+:=s01@F3s;"G-,5.2;"*-3g4I?f9<=5.1D4$.1*:@<3F;"+:=A0V4?A9<,58:;eB:@uB:@7GZ9<*H4$02.13/4I,5.2;$*W +:*:@78:@<+:=5.13\,5.2^6+:@74$G:GH4$=54$.1,GZ;"+:=A0V4?f4$*:.1>F=\@B:@uc4$.1=5@[I4$=\.2@<=C35.1?j+:01,/4$*:9<?A@F*6,E02@<3CGH4$=/4I?f>X,/=5@<3

(N,M,K)@X,j01@fD<;@ADF.1@F*6,jB:@fGw9<*H4$01.2354I,/.1;"*

n¡.10

DF;$*6[e.2@<*:B:=/4I.1,B-@O?f;"*R,/=\@F=,58:9F;$=5.2^6+:@<?f@<*6,DF@X,5,/@O=5><`"02@IH;"+-=0V4j?A9<,/8-;eB:@OB:@O=\9F`"+-0V4$=\.2354I,/.1;$*W

Page 203: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

lj 7 MP

0V4*-9FDF@<3535.k,/9B-@scd;"*:DX,/.2;$*:3jB-@g=59<`"+:0V4I=5.2354I,/.k;"*φn

4$BH4$G,/9F@<3A4I+ G:=\;"b:01>F?A@mB:.23\DF=59X,/.1359s35@sc4$.k,35@<*6,5.2=F

;"+,\;"+-=53jGZ;$+:=a02@<3a;"G,/.2;$*:3a4$?f9<=5.1D4$.1*:@<3FW*:;"+:3j4y[";"*-3ac4$.k,[email protected]@F* @F*6,5=5@s*-;$,/=\@s4$02`$;"=5.k,58:?A@@<,0V4?A9<,58:;eB:@OB:@C ;"*:`"3M,4IvY_MeD 84$=O?m4$.13.1035@F=54$.1,9F`R4I02@F?A@F*R,.2*6,59F=\@F35354$*6,B:@O[";".1=D<@C^6+:@B:;"*-*:@f0rqt+-,/.101.1354I,/.k;"*B:@ ∑

zN,R,Mk (PNtk

)∆Wk

DF;$?f?A@?m4I=\,/.1*:`R4I01@D4I*:B:.2B:4I,/@ u024u?A9X,/8:;eB:@B:+H4I02@QB:9<DF=5.k,/@OG:4$=;"`"@<=53B:4$*:3EL';"`R) S

;$+:34[";"*:302.1?A.k,59*:;"3,5@F3\,53*e+:?A9F=\.2^6+:@<34$+AD4$3Bq'+-*s4$0194Ob-=5;*-.2@F*f@X,+:*:@`$9F*:9<=/4$01.2354I,/.1;"*Bxq 4$+,/=5@<34$0194$33\@F=54$.1,.1*6,/9<=5@<353/4I*6,/@I

0HG:4$=/4$.k,.2*R,/9<=5@<353/4$*R,B:@cd;"=\DF@F=0rq'4$G:G-=5;yp.1?m4T,/.k;"* Q<,/=\@+:*:@cd;"*:DX,/.2;$*A .1G:35D8:.1,GZ;"+:=`R4$`"*-@F=+:* c4$D<,5@F+:=

hJ [";".2=024mB:.23\DF+:3\35.1;"*G:4$`"@ % #"#"U@X,QB:;"*:DjB:@?A;eB:.1]:@F=0rq'4$02`$;"=\.k,58:?A@CGZ;"+:=,5@F*:.1=

DF;$?fG-,5@OB-@ED<@<,4$35GZ@<D<,C 0Hc4I+:B:=/4I.1,9<,5+:B:.1@F=3\.Y0rq'4$02`$;"=\.k,58:?A@35@`$9F*:9<=/4$01.23\@4$+fDF4$3B:@F3 QG:=5;"`$=5@F3\35.k["@F3=59X,/=5;$`"=/4$B-@F3F 0H4$G:G:4$=/4$.k,9<`R4$01@F?A@F*R, ,/@<*6,4I*6,Bqt+-,/.102.135@<= B-.2=5@<D<,5@F?A@F*R,B:@F38:.13\,/;$=5.2^6+:@<3Bqt9<[";$02+-,5.2;"*B:@

PN

0V4CG:024$D<@B:@F335.1?+-0V4I,5.1;$*:3B-@O7;"*R,/@X_4$=501;gJ9<[e.k,4$*R,4$.2*-35.w0qt@F3\,5.2?f4I,/.k;"*fBqt+:*g?f;eB->F01@yUX-qt@F3M,0rqt.2B-9F@Q@ pGZ;"359<@CBH4$*:3EL') % S^6+qt.20DF;$*6[e.2@<*:B:=/4I.1,B:@Q["9<=5.k]H@F=,/8-9F;"=\.2^6+:@<?f@<*6,

)

Page 204: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

%(' ) ) $)+!# "# % '*)+ *)+

! #"%$& ')(%"*$'+!,;".k,

T > 00rqt8:;"=\. F;$* B:+KG:=5;"b-02>F?A@$h*KDF;"*-35.2B->F=5@m+-* @F3\GH4$D<@sB:@mG:=\;"bH4$b-.202.,/9

(Ω,F ,P)^6+:.

35+-G:GZ;"=\,5@+:* ?f;"+["@F?A@F*R,mb:=5; *:.1@F*W

B:@7B:.2?A@F*-35.2;$*q

h* =/4IG:GZ@F0102@u^6+:@<02^6+:@<3s*:;$,/4I,/.1;"*:3DF024$353\.2^6+:@<3B:@<3 Q- Jd[";".1=OLt N-PQ&RTSVU¡ (Ft)0≤t≤T

B:9<35.2`$*:@C0V4]H0k,/=54I,/.1;"*@F*:`$@F*:B-=59F@aGH4$=W

@<,4I+:`"?A@F*R,/9F@Ixh**-;$,/@ P 0V4,5=5.2b-+7B:@<335;$+:3M_@<*:35@<?jb:02@<3G:=59X[e.235.1b:01@F3B:@

Ω × [0, T ]

L2t (R

q)B-9F35.1`"*:@g0rqt@F*:3\@F?jb:01@gB:@<3[T4$=5.24$b:02@<34$029F4I,/;$.2=\@F3 [T4$02@<+:=53

RqW Ft

?f@<35+:=54$b:01@F3@<,B:@D4I=5=59O.1*6,59F`"=54$b:01@$Hh*u*:;$,[email protected]?fG:01@F?A@<*6,

L2t

GZ;"+-=L2t (R)

H2

T (Rq)B:9F3\.2`"*-@g0q'@<*:35@<?jb:02@mB:@<3G-=5;eDF@<353\+:3AG-=59<[e.135.2b-02@F3

φ : Ω × [0, T ] → RqJD<;"*:3\.2B:9<=59<3

DF;$?f?A@B:@<3["@<D<,5@F+:=\301.2`"*-@F3U ,/@F013^6+:@E∫ T

0|φt|2dt

35;".k,]H*:.:h*s*:;$,5@3\.2?AG:02@<?A@F*R,H2T

GZ;"+:=H2T (R)

-;"+:=

β > 0@<,φ ∈ H2

T (Rq)W ‖φ‖2

β

B-9F35.1`"*:@E∫ T

0exp(βt)|φt|2dt

H2T,β(R

q)B-9F35.1`"*:@0q'@<35GH4IDF@

H2T (Rq)

?j+:*:.B:@O024*:;"=\?f@ ‖ · ‖β

9F35;$+:B:=5@O+-*:@O9<^e+:4I,/.1;"*B:.kvx9<=5@F*R,/.1@F0102@3\,/;eD8H4$3M,/.1^e+-@O=59<,5=5;"`$=/4$B:@DF;"*-35.23M,/@ j,/=\;"+-["@F=+:*DF;"+:G-02@B:@OG:=\;eDF@F3\35+:3

(Y, Z)4$BH4IG-,/9"

(Ft)0≤t≤T^6+:.["9F=\.1]H@¡

YT = ξ,

−dYt = f(t, Yt, Zt)dt− ZtdWt

J/.a % U

;10ξ

@F3M,+:*:@[T4$=5.24$b:01@4$0194T,/;".1=5@ FT

_l?f@<35+:=54$b:02@@<,f : Ω×R+×R×Rq → R

@<3\, P⊗B⊗Bq _?A@F35+-=/4$b:01@$f

@F3M,f4$G:GZ@<029g01@gB:=\.1["@<=B:@0q QE ;"=\35^6+:@ξ

4$G:GH4$=M,/.1@F*6,L2T

W ^6+:@f(·, 0, 0)4$G:G:4$=\,5.2@F*R,

H2T

@<,^6+:@f

@F3M,+:*:.kcd;"=5?A9<?f@<*6, .1G:35D8:.1,O4$+u3\@F*:3;10 ¡

∃C > 0, dP× dt p.s.|f(ω, t, y1, z1)− f(ω, t, y2, z2)| ≤ C(|y1 − y2|+ |z1 − z2|)∀(y1, z1, y2, z2),;"*uB-.1,^6+:@

(f, ξ)@F3\,+:*uDF;$+:G:02@QB:@OGH4$=54$?f>X,/=\@F33M,4$*-BH4$=5Bx

h*74A4I02;"=\3EJ [";".2=QL2--&")IS;$+ L' N3-PQ&RTSVU¡465879;:=<?>@BA DCFE=GIH

(f, ξ) J?KML E JONQPSR6T?RUNWVYXVYZ\[ H X]R]^_^ H VYKQT`VYXT=a G PbRc G ^ H RJ!KdJ?K Gfe JgRhL E JONQPSRT?RiNQX E L]R]^]^jJ`^ (Y, Z) ∈ H2

T × H2T (Rq)

e J G Xlk]^ E J HaJf.nm % Upo

Page 205: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.

4aG:=5;$GZ;"35.k,/.2;$*m3\+:.1[T4$*6,5@J\LN3-PO&TS UGZ@<=5?A@<,Bqt@XpG:=\.2?A@F=@ p-G-02.2D<.1,5@F?A@F*R,Y

02;"=\35^6+:@02@B:=5.k["@F=f

@F3\,02.1*:94I.2=\@@<*(y, z)

m :=9 9 /9 A CFE=GIH

(δ, γ) J!K NQX E L]R]^]^jJ`^NQX]k G ^ G PSR]a jE XK'k]a=VYPSRjJ?Xl^ (R,Rq) a φ J?K kjPSkZ\RjK H T?R H2

T (R) R H ξ ∈ L2T

o P E Xl^P C"!

YT = ξ,

−dYt = [φt + Ytδt + Ztγt]dt− ZtdWt

V\J?K'RJ?K Gfe JgR ^ E P J HfG E K (Y, Z) T`VYK ^ H2T,β(R) × H2

T,β(Rq) R H Y R]^ H T E KgK'kiN VYX3P V$# E XZhJ?PSR

Yt = E(ξΓtT +

∫ T

t

Γtsφsds|Ft

)

E&%Γts R]^ H PSRiNQX E L]R]^]^jJ`^nV`T(' E=G K H T?k)K G N E J?X s ≥ t N VYX P * C # E X,+ VYXT

dΓts = Γts[δsds+ γ∗sdWs],

Γtt = 1.

KN VYX H G LjJ?P G RjXla^ G ξ R H φ ^ E K H N E ^ GIHfG #l^]a PSRNQX E L]R]^]^jJ^ Y R]^ H N E ^ GIHfG # o C;G T?RUN&P J^]a Y0 = 0 aiVYP E Xl^N E J?X HE J H t a Yt = 0 N o ^ o a ξ = 0 N o ^ o R H φt = 0 dP ⊗ dt N o ^ o

.- /10"2430 ,' ,#65_"* #87% # :9<;=>@?:G-GABACDB^E?"FHGqI J>@K J/.6mMLoUG"NFOC"C"FHGQPacANRSFQG"PTN(UV2DO?"GWVYXZN(FB?"Vk[[PTC\U]F¡

Xt =X0 +

∫ t

0

b(s,Xs)ds+

∫ t

0

σ(s,Xs)dWs,J/.6m_^"U

Yt =Φ(X) +

∫ T

t

f(s,Xs, Ys, Zs)ds−∫ T

t

ZsdWs

A10b, σ, f,Φ

[`ONV1]"FZCEU$GqIaRnGA[U]a"XOBFA¡

cb 9dO58<e@ A gf R]^# E K'L H G E K ^ b a σ R H f ^ E K Hihkj P T?kjX G RjKgK'R]^ Tl G KQT G LR 12RjK H RjZ%Ng^ R H J!K G

# E XZ\kjZ\RjK Hf G Ng^ Lnm GIHWo T`VYK ^PSR]^=VYX G V PSR]^ Tl R]^ N V?L]R oHp K ^jJON`N E ^ R e JgR Φ : Dd → R R]^ H J?K'R# E K'L HfG E KgK'RjPIPSR L∞ f G Ng^ Lnm GIHWo VYJ ^ RjK ^ E[% a ∀x1, x2 ∈ Dd

|Φ(x1) − Φ(x2)| ≤ C sup0≤t≤T

|x1(t) − x2(t)|

VTYR]L D P R]^ N V?L]R T?R]^i# E K'L H G E K ^ L E K H G KgJgR]^qBTYX E=GIH R]a VTYR]L T!R]^ P G Z GIH R]^qsr`VYJgLnmgRtL`T PMru aUT?k)K G R]^ ^jJ?X [0, T ]

ovp K K E=H RhN VYX K > 0HE J H R]^BPSR]^dL E K ^ H VYK H R]^ T?R f G Ng^ Lnm GIHQo R H E K ^jJON`N E ^ Re JgR

sup0≤t≤T

|b(t, 0)| + |σ(t, 0)| + |f(t, 0, 0, 0)| + |Φ(0)| ≤ K

E&%0 R]^ H P V$# E K'L H G E K L E K ^ H VYK H Rw=VYP VYK H 0 ^jJ?X [0, T ]

o^Tx$|

Page 206: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.6mML1m J>DK

@?>W:8@ A tm V .u L VYJ`^ RBT?R P VXlRjZ VYX e J;R o?o MT`VYK ^m V la N E ^]^ [ T?R J?K'RYRjXl^ G E K L`T PMr ^ G a RjK NQP J`^nT?R P m ]N E H mg[]^ R no =a σ R]^ H J?K G # E XZ\kjZ\RjK H f G Ng^ Lnm GIHWo N VYX XVjN`N E X H H o

465879;:=<?>@BA tm V .u p K ^jJ N`N E ^ R e JgR P m N E=H mg[]^ R no nR]^ H YkjX G )ik]RUR HFe J;R Z R]^ H L`T PMr oCFE=GIHπ = (ti)0≤i≤n J!K'R6T G ^ LjXlk HfG ^V H G E K T?R [0, T ] a VYP E Xl^

n∑

i=1

E

∫ ti

ti−1

|Zt − Zti−1|2 + |Zt − Zti |2dt ≤ C(1 + |X0|2)|π|

E&%C > 0 R]^ H J?K'RhL E K ^ H VYK H R T?k NgRjKQT`VYK H T?R T R H K Z V G ^ G KQT?k NgRjKQT`VYK H RnT?RnP V T G ^ LjX]k HfG ^V HfG E K

πo

%(' * )+ $) $)+ " $%('*)+ )+ *)+

4CsC"A?DBiP!"!"?n\PTCEU B?"N$#NON-+ % oS'&)("ATCEU AC NF!"NFOC"(sGWFZB*CDATUnP&U]VWATC"BUXW?"C"F `<^e?"P[U]VYAC*("Vkv `ONFZCe_

U]VYFZGYGYFB(UA +,adP[B(U]V1^E?"FHN(`OU]N(A"`NP("FJN`,"`+na"VWF*& J>DK$KUFOB(U-("`<]dCDVWF.!dPTN GQP/("ACDC"`ZF0(DF&VVU ?"C"F.+ZATC"("VYUVWAC UFZNRSVWC"PTGYF

ξ ∈ L2W

&VYV Us?DC"FcAC1+OU]VYACf : Ω× [0, T ]×R×Rq → R

W\U]FOGWGYF^E?"F ∀(y, z) ∈ R×RqWf(·, y, z) ∈ H2

T

FUU]FOGWGYF^e?xqIVYGEFXp@VWB U]F

C > 0B]P&U]VWBMcPTVWBPTCEU |f(t, y, z)−f(t, y′, z′)| ≤ C(|y−y′|+|z−z′|) ∀y, y′ ∈

R, z, z′ ∈ Rq, p.s.W

&VYVWV2Us?"C1A"bDB(UnP2+ZGWF(St)0≤t≤T

WZ^e?DV FZB(U?"C3!"NA +OFZBB(?"B A[[PTGYFZ?"N(BN`OFZGYGWFOBFW4+OACEUVWCe? FU5!"N(A"`NFOBB(V1[F _RSFZCEU RSFZB(?"N]P$bDGWFJB]P[UVWBMcPTVWBPTCEU

E(sup0≤t≤T (S+t )2) <∞ mZhCcB?1!"!ABFE^E?"F

ST ≤ ξ p.s.6 PsB(AGW?@U]VWATC3(DF Gq_J>DK$K7("`X]dC"VWF8!dP[N(ξ, f, S)

FZB UJ?"C U]N(V!"GYFOU(Yt, Zt, Kt)0≤t≤T

("F/!DNA +ZFOBB(?"BFt

!"NA"`TNFZB(BVk[FZRSFZC\U RSFZB(?"N]P$bDGWFZB+[&PTGWFO?"NBR

WRq

FUR+

FOU$B]P&U]VWBMcPTVYB]PTC\U&Vk[-U

Z ∈ H2T (Rq) 9&d[-U

YFZB U ?"C!"N(A +ZFZB(B?"B:!"N(`<[eVWB(V2b"GYFUFZG ^E?"F

E(sup0≤t≤T |Yt|2) <∞ 9&d[eVVUYt = ξ +

∫ T

tf(s, Ys, Zs)ds+KT −Kt −

∫ T

tZsdWs

W0 ≤ t ≤ T 9&d[eVYV U

Yt ≥ StW0 ≤ t ≤ T 9&d[eVYVWVVU

(Kt)0≤t≤TFOB(U ?"C.!"NA +OFZBB(?"B;+ZAC\U]VYCe?W+ZN(AVWB(B]PTC\UWZCe?"GTFZC

0FOU UFZG"^E?"F ∫ T

0(Yt − St)dKt = 0

m6 FJU]a"`OANXORSF=<!m_^>(dP[C"B=#NON3-

+ % oS?!FONRSFOU P[GWAN(B@(qIP SNRSFON^E?"F

465879;:=<?>@BA f C"! ! T?k)K G RN VYXst G u&,t GIG u&,t GIGIG uV`TYZ\R H J!K'R6J!K Gfe JgRn^ E P J HfG E K (Y, Z,K)

^V HfG ^#pV G ^VYK H t G u&,t.u&,t. G u&,t. GIG u ,t. GIGIG u o^TxA<

Page 207: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

. >

AT?"BPTGYGWATC"B!"N`OBFZC\U+ZATC"BV("`ZN(FZN$("FOB >DK K RsP[N5TAy[@VYFZC"CDFZBm 4C !dPTN U]V+Z?"GYVYFZNFWdATC +OAC"B(V("XONFGQP/(-neCdPTRSV2^E?"F&f.nm ^"U !A?"N

X&ACcC"ATUF.!dPTN

XNBACcB +na"`ORsP (qI4?"GWFONUFU$GWFOB ("FO?p J>DK$K¡

Yt = φ(XT ) +

∫ T

t

f(s,Xs, Ys, Zs)ds−∫ T

t

ZsdWs +KT −Kt,

Yt ≥ g(t,Xt), t ≥ 0,

∫ T

0

(Yt − g(t,Xt))dKt = 0,

Py[TF+φ(·) ≥ g(T, ·) FOU

Y N,0t = φ(XN

T ) +

∫ T

t

f(s,XNs , Y

N,0s , ZN,0

s )ds−∫ T

t

ZN,0s dWs +KN,0

T −KN,0t ,

Y N,0t ≥ g(t,XN

t ), t ≥ 0,

∫ T

0

(Y N,0t − g(t,XN

t ))dKN,0t = 0.

&f.6mRU

6 FZB aRn !ATU]a"XOBFOB5("F*#x <TS BAC\UGYFZB$B(?"V1[&PTCEUFZB/&XFOUW

B(ACEU B?1!"!AB`OB5(DF=("VYR FOC"BVYACd

WY^E?"VFZB U FOC"B?DVYU]FQ]-p@`ZF

1(dP[C"B=#xA<oS Py[F+H?"C"FwFXpeU]FOC"BVYACucP+OVWGYFJPT?+ZPTB

d > 1U¡

cb 9dO58<e@ A b R H σ ^ E K H L E K H G KgJgR]^]a T G kjXlRjK HfG V PSR]^ R H VTYR]L%T!R]^%T?kjX G Yk]R]^ J?K G # E XZ\kjZ\RjK HjE XK'k]R]^iN VYX3XVjN`N E X H \P Vs=VYX G V PSR x aUR H σσ∗ ≥ ε0Id

E&%ε0 > 0

ocb 9dO58<e@ A

f R H φ ^ E K H L E K HfG KgJgR]^ R H J!K G # E XZ\kjZ\RjK H$f G Ng^ Lnm GIHQo N VYX\XVjN`N E X H VYJ c =VTX G V PSR]^ (x, y, z)

ocb 9dO58<e@ A

g ∈ C1,2([0, T ] × Rd) VTYR]L H E J H R]^ ^ R]^nT!kjX G Yk]R]^ J!K G # E XZ\kjZ\RjK H jE XK'k]R]^ o @?>W:8@ A nVYK ^$ 3 la P ET ^ H V?LjPSRhR]^ H K E=H k h(·, ·) oE J`^hVT E K ^NQXlk# kjX]knZ E T G )ikPSR]^K E=H V H G E K ^*L VYXT`VYK ^iK E=H X]R*L V=^]a G P=N E J!XXV GIH nVT E=G XiJ?K'R%L E K #J`^ G E KVTYR]LiPSRbN V=^T?R H RjZ%Ng^ h = T

N

oiVYUAC"B !"N(`ZBFOCEUHGWF GYFZRSR Fv^?m ^ ("F #x <TSW^E?"FqC"AT?"Bw`+ZNVk[AC"B(dPTC"BJ?"C + P(DNFqR ?DGYU]V(DVYRSFOC_BVYAC"C"FOG¡ @!>>@BA p K ^jJON`N E ^ R e JgRPSR]^m ]N E=H mg[]^ R]^ o a o R Ho B^ E K H XV G R]^ o

o P Rc G ^ H R3J?K'R L E K ^ H VYK H R3J!K G YRjX]^ RjPIPSR C > 0H RjPIP R e J;R

E sup0≤t≤T

|Yt|2 +

∫ T

0

|Zt|2dt+K2T ≤ C(1 + |X0|2)

E sup0≤t≤T

|Y N,0t |2 +

∫ T

0

|ZN,0t |2dt+ |KN,0

T |2 ≤ C(1 + |X0|2).

?o

E

sup0≤t≤T

[|Y N,0t − Yt|2 + |KN,0

t −Kt|2] +

∫ T

0

|ZN,0t − Zt|2dt

≤ C(1 + |X0|2)√h.

^Tx

Page 208: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.nm ^!m J>DK$K

o p K K E=H RUN E J?X t ∈ [tk, tk+1[

kt =|∂tg(t,Xt)| + |d∑

i=1

∂xig(t,Xt)bi(t,Xt)| +

1

2|

d∑

i,j=1

(σσ∗)i,j(t,Xt)∂2xi,xj

g(t,Xt)|

+ |f(t,Xt, Yt, Zt)|,

kN,0t =|∂tg(t,XNt )| + |

d∑

i=1

∂xig(t,XN

t )bi(tk, XNtk

)| + 1

2|

d∑

i,j=1

(σσ∗)i,j(tk, XNtk

)∂2xi,xj

g(t,XNt )|

+ |f(t,XNt , Y

N,0t , ZN,0

t )|. P E Xl^]a VYJ ^ RjK ^nT?R]^3ZR]^J!XlR]^nVYPSk V HE=G XlR]^]a E K V

0 ≤ dKt ≤ ktdt, 0 ≤ dKN,0t ≤ kN,0t dt,∀t ∈ [0, T ], p.s.

iFHGYFZRSRSFBFqIP!"!"?DVWFwB?DN-(DFZB$N(`ZB?DGYUnP&U]B ("F#NEN3-+ % oS FU FOC!dPTN(UV+O?"GWVYFZN

:=9 9 /9 A p K ^jJ N`N E ^ R e JgR P ET ^ H V?LjPSR g(t,Xt) R]^ H J?K'R%^ RjZ G Z VYX HfG Ker`VYPSR T?R_P V # E XZ\R

g(t,Xt) = g(0, X0) +

∫ t

0

Usds+

∫ t

0

VsdWs

E&%U R H V ^ E K H T?R]^iNQX E L]R]^]^jJ^ Ft SNQX E rYX]R]^]^ G YRjZ RjK H Z\R]^jJ!XV PSR]^%^V H G ^#pV G ^VYK H

∫ T

0

(|Ut| + |Vt|2)dt <∞ p.s.

P E X]^Zt = Vt, dP × dt p.p. ^jJ?XP RjK ^ RjZ P R Yt = g(t,Xt)

&f.6mt|6U

R H

0 ≤ dKt ≤ 1Yt=g(t,Xt)[f(t,Xt, g(t,Xt), Vt) + Ut]−dt.

hCc?DUVWGWVMBFZNP `F`\P[GWFZRSFOCEU ?"CDFHN(FZR PTN5^E?"FwGYVW`OF >+ZFwN`ZB(?"GYU]P[UO¡

@?>W:8@ A f V NQX E NQX G k H kML G T?R]^]^jJ`^NgRjJ HjH X]R r?kjK'kjXVYP G ^ k]R #pV?L G PSRjZ\RjK H J?K ET ^ H V?LjPSRg(t,Xt)

e J G R]^ H J?K'R ^ RjZ G Z VYX H G Ker`VYPSR NQP J^ r?kjK'kjX VYPSR

g(t,Xt) = g(0, X0) +

∫ t

0

Usds+

∫ t

0

VsdWs + At

^Tx\

Page 209: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

. >

E&%A R]^ H J?K NQX E LR]^l^jJ^hL E K H G KgJ =VYX G V HfG E K G K H k rYXV PSR H RjP e JgR6P V Z\R]^jJ?XlR dAt R]^ H ^ G KerYJ?P G [jXlR

N VYXiXVjN`N E X H dt R H e J G V`TYZ\R H J?K'R T?k]L E Z%N E ^ GIH G E K At = A+t −A−

t a E&% A+t R H A−

t ^ E K H T?R]^bNQX E L]R]^]^jJ`^ G K H k rYXV PSR]^ o P E X]^]a (Ut)t R H (Vt)t ^ E K H T?R]^NQX E L]R]^]^jJ`^ Ft NQX E rYXlR]^]^ G YRjZ\RjK H Z\R]^jJ?X V PSR]^XlR]^ NgR]L H G YRjZ\RjK H s=VYPSRjJ?X]^nT`VYK ^ R R H Rd ^V H G ^#pV G ^VYK H

∫ T

0

|Ut| + |Vt|2dt+ A+T + A−

T <∞ p.s.

f k r`VYP GIH k &/.6mt|6U R]^ H HE J,' E J!Xl^%^V HfG ^#pV GIH RnR H PSR XlR]^ H R6R]^ H Z E T G )ik N VYX tVTYR]L Lt PSR H RjZ%Ng^_P E L VYP8RjK0 T?RP Vh^ RjZ G Z VYX H G Ker`VYPSR Yt − g(t,Xt) u

dKt +1

2dLt =1Yt=g(t,Xt)[f(t,Xt, g(t,Xt), Vt) + Ut]

−dt+ dA−t .

P RjKdT?k]L E J!PSR P Rc G ^ H RjK'L]RTl J!K NQX E L]R]^]^jJ`^iNQXlk G ^ G PSR (αt, 0 ≤ t ≤ T )H RjP e JgR 0 ≤ αt ≤ 1 R H

dKt = αt1Yt=g(t,Xt)([f(t,Xt, g(t,Xt), Vt) + Ut]−dt+ dA−

t ).

6 FZB$PT?DU]FO?"NB ("F#xA<oS?("`X]dC"VYBBFOCEUPTGYANB?"C"FHP!"!"N(Ayp@VWR P[UVYAC (DF &/.6mRU !dPTN hCcVWCDVYU]VWP[GYVMBFFOC

tN!dP[N

Y N,1tN

= φ(XNtN

)m

-4?"VWB!A?"Nt ∈ [tk, tk+1[

WZACwN(`ZB(A?DU Grq`F^E?dP[UVWACN`OUNA"`TN]P("FC"ACwN(`,d`+na"VWF ("F4BAGY?DU]VYAC(Y N,1, ZN,1)

¡

Y N,1t = Y N,1

tk+1+

∫ tk+1

t

f(s,XNs , Y

N,1s , ZN,1

s )ds−∫ tk+1

t

ZN,1s dWs.

hC("`X]dC"VYU FZC"B(?"VYUF.!A?"N UA?DUt("PTC"B

[tk, tk+1[WY N,1t = Y N,1

t ∨ g(t,XNt )m

hCcVYUXZNF (?DB5^E?q qGQP/(dP[U]Ft0 = 0

m G FZB(UcP+OVWGYF=("FC[AVWN^E?"F

Y N,1tk

= uN,1k (XNtk

)!AT?"N?DC"F/+ZFON(UnP[VWC"FCcAC"+U]VYAC1RSFOB?"N]PIb"GWF

uN,1k

m 6 FGWFOR RSF !m ^/(dPTC"B0#xA<oS?("ATC"C"F.("FZB +ZAC\U]N TGWFZB B?DN

(uN,1k )kWY N,1

FUZN,1

m @!>>@BA p K ^jJON`N E ^ R3PSR]^m ]N E=H mg[]^ R]^ o a o R Ho B^ E K H YkjX G )ik]R]^ o f R]^ # E K'L HfG E K ^ uN,1k ^ E K H4f G Ng^ Lnm GIHQo J?K G # E XjZ\kjZ\RjK H RjK N R H k o P Rc G ^ H RJ?K'R L E K ^ H VYK H R C a G KQT?k NgRjKQT`VYK H RT?R N a H RjP PSR e JgR ∀k VYKQT ∀t ∈ [tk, tk+1[

E

suptk≤t≤tk+1

[|Y N,1t |2 + |ZN,1

t |2]

≤ C(1 + |X0|2)

E|Y N,1t − Y N,1

t |2≤ C(1 + |X0|2)|tk+1 − t|.

6 F Ua"`ZATNXZRSF m:("PTC"B #xA<oSbANC"F GrqFZN(NFZ?DN FOCEUNF(Y N,0, ZN,0, KN,0)

FU(Y N,1, ZN,1, KN,1)

¡465879;:=<?>@BA p K ^jJ NN E ^ R e J;RPSR]^m ]N E H mg[]^ R]^ o a o R Ho B^ E K H XV G R]^ o P E X]^

E

sup0≤t≤T

[|Y N,1t − Y N,0

t |2 + |KN,1t −KN,0

t |2]+

∫ T

0

|ZN,1t − ZN,0

t |2dt

≤ C(1 + |X0|4)h.

^Tx z

Page 210: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.6m m . 6 *> 6 .>"> P >

4C-]dCWGWFJU]aD`ZAN(XZRSF m(dP[C"B8#xA<oS;!FONRSFOU:(xqI`X[[PTGY?"FZN$GQPvN`<`?"GQP[NVYU`L2

("FZm iFwU]a"`OANXOR F

B?1!"!ABFE^E?"Fcb 9dO58<e@ A

b a σ R H f ^ E K H J?K G # E XZ\kjZ\RjK Hhkj P T?kjX G RjKgK'R]^Tl G KQT G L]R 12RjK H RjZ%Ng^ o

465879;:=<?>@BA p K ^jJ N`N E ^ R e JgRhPSR]^km N E=H m;[]^ R]^ o a o a no )R Hio ^ E K H XV G R]^ o P E Xl^G PbRc G ^ H R3J!K'R L E K ^ H VYK H R C > 0H RjPIPSR e J;R

N−1∑

k=0

sups∈[tk,tk+1]

E

∫ tk+1

tk

|Zt − Zs|2dt ≤ C(1 + |X0|4)√h.

U #$%(')+

C+OFZN(U]PTVWC:C"ATRbDNF(qVWC"`<`\PTGYVYU`ZBBATCEU-+ZATC"B(U]PTR RSFOCEU ?DU]VYGWVYB(`ZFOBm 7 7 8@39 l

∀ γ > 0,∀ (a, b) ∈ R2, (a+ b)2 ≤ (1 + γh)a2 + (1 +1

γh)b2.

&/.6m <"U @/@!>>@ 8@ :=9/l 1:O@\

@!>>@BA CFE=G RjK H T > 0 R H tk = kh N E J!X 0 ≤ k ≤ N = ThJ?K'R T G ^ LjXlk H G ^V HfG E K T?R

[0, T ] T?R_N V=^ h oE J!X HE J H R]^ ^jJ GIH R]^_N E ^ GIHfG YR]^ (ak)0≤k≤N , (bk)0≤k≤N R H (ck)0≤k≤NH RjPIPSR]^ e JgR

ak−1 + ck−1 ≤ (1 + γh)ak + bk−1 t.+ GIH m γ > 0 u a E KdV

ak +N−1∑

i=k

ci ≤ eγ(T−tk)[aN +

N−1∑

i=k

bi].

&/.6mRU

U )+# $%('*)

AT?"B GYVWB U]AC"BwV+OV ?"C +OFZN UnPTVYC C"ATRbDNF("F (D`<]dC"VMU]VYAC"BHFOU N(`ZB?DGYUnP&U]BkF p1!ATB`ZB/(dP[C"B #ENON x\^yS mhC !FO?DUk`F`PTGWFOR FOCEU.+ZATC"B?"GMU]FONHB?DN0+ZFSB? (FOU # - AG z |ISW # 6 % LFSW # ! % IS A? # ?1( % % S m AT?"BNFOR FON +OVWACDB nmx*Pn\P[U]VYB!A?DNwGWPsU]NP("?"+U]VYACicNPTC#"ZPTVWB(F>("FZBJUFZNRSFZBJGWVY`ZBJP[?p C"ARjb"N(FZB("F +OA?_[FZN U]?"N(FYm

%$ & 2('*) 0W #67* $ 2 (,+ 0?!( 0W - 7/.l /9 A CFE=G RjK H ε > 0 a G J!K'R LjP V=^]^ R T?R # E K'L HfG E K ^ Rd → R a 1 ≤ p < ∞ a R H ν J?K'RZ\R]^jJ?X]RnT?RiNQX ET V G P GIH k_^jJ?X Rd

oE J?XJ?K'R # E K'L HfG E K g : Rd → R a E KdT?k)K GIH

‖g‖Lp(ν) := ∫

|g(z)|pdν 1p .

^Tx %

Page 211: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

. >

oQE J H R #pVYZ G PIPSRi)K G R T?R# E K'L HfG E K ^ g1, · · · , gN : Rd → RH RjPIPSR e J;R%N E J!X HE J H R # E K'L HfG E K

g ∈ G a G P Rc G ^ H R j = j(g) ∈ 1, · · · , N H RjP e JgR‖g − gj‖Lp(ν) < ε

R]^ H VjN`NgRjPSk]R J!K'R ε L E JeYRjX H J?XlR6T?R G N VYX XVpN`N E X H ‖ · ‖Lp(ν)

o?o p K K E=H R N (ε,G, ‖ · ‖Lp(ν)) P V H V G PIPSR6T?RP VNQP J`^UNgR H GIH R ε L E JeYRjX H J?XlR6T?R G N VYX3XVjN`N E X H

‖ · ‖Lp(ν)

op K NQXlRjKQT N (ε,G, ‖ · ‖Lp(ν)) = ∞ ^d G P K Rc G ^ H R N V=^\T?R ε L E JeYRjX H J!XlR o4 P E Xl^N (ε,G, ‖ · ‖Lp(ν)) R]^ H VjN`NgRjPSk3K E Z XlRnT?R ε L E J@YRjX H J!XlR6T?R G N VYX3XVjN`N E X H ‖ · ‖Lp(ν)

o o3CFE=GIH zM1 = (z1, · · · , zM ) M N E=G K H ^ ) c!R]^ T?R Rd

obCFE=GIHνM P VZ\R]^jJ!XlRhRjZ%N G X Gfe JgRhL E XXlR]^

N E KQT`VYK H R]a Gfo R o

νn(A) =1

M

M∑

m=1

1A(zm), (A ⊂ Rd).

P E Xl^

‖g‖Lp(νM ) = 1

M

M∑

m=1

|g(zm)|p 1p

R H HE J H R ε L E JeYRjX H J?X]R T?R G N VYXX VjN`N E X H ‖ · ‖Lp(νM ) ^ RjXV VjN`NgRjPSkJ?K'R Lp ε L E JeYRjX H J?XlRT?R G ^jJ?X zM1 R H P RK E Z XlRnT?R ε L E JeYRjX H J?XlR6T?R G NWVYX XVjN`N E X H ‖ · ‖Lp(νM ) ^ RjXV\K E=H k

Np(ε,G, zM1 ).

K Tl VYJ H XlR]^ H RjXjZ\R]^]a Np(ε,G, zM1 ) R]^ H P RjK H G RjX N Z G K G Z VYP H RjP e J G P&Rc G ^ H R*T?R]^# E K'L H G E K ^g1, · · · , gN : Rd → R VTYRLhP V NQX E NQX G k H k e J;R N E J?X H E J H R g ∈ G a G P Rc G ^ H R j = j(g) ∈1, · · · , N H RjP e JgR

1

M

M∑

m=1

|g(zm) − gj(zm)|p 1p < ε.

%$ .- 2U $' #" !87f,#$ ! ' #" ' 7% ' #"*! 2"

- 7/.l /9 A CFE=GIH A J!K'R LjP V ^l^ RnT?R ^ E J`^ RjK ^ RjZ PSR]^3T?R Rd R H ^ E=GIH M ∈ No

o E J?X z1, · · · , zM ∈ Rd a E K T?k)K GIH

s(A, z1, · · · , zM) =∣∣A ∩ z1, · · · , zM : A ∈ A

∣∣,

Ld R]^ H TT G XlR e JgR s(A, z1, · · · , zM) R]^ H PSR K E Z X]R T?Rd^ E J`^ RjK ^ RjZ PSR]^ T G ^ H G K'L H ^MT?Rz1, · · · , zM RjK G K H RjXl^ R]L H VYK H VTYR]LnT?R]^ kjPSkjZ\RjK H ^nT?R A o

^@Lx

Page 212: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.nm'| m K,wK$ >D> h h _ -&.K . K P

?o3CFE=GIHG J?K ^ E J`^ RjK ^ RjZ PSRMT?R Rd T?R L VYXT G KQVYP M ocp K T GIH e JgR A #XV rYZ\RjK H R G ^ G

s(A, G) = 2M a Gfo R o ^ G H E J H ^ E J`^ RjK ^ RjZ PSR\T?R G NgRjJ H jH X]R X]R NQXlk]^ RjK H kh^ E J^ P V # E XZ\RA ∩G N E J!X3J?KkjPSkjZ\RjK H A ∈ A o

oJf R [jZ\R L E R L G RjK H Tl k]LjP V H RjZ\RjK H S(A,M) T?R A R]^ H T?k)K G L E Z Z\RS(A,M) = max

z1,··· ,zM⊂Rds(A, z1, · · · , zM).

*G K ^ G aUPSR [jZ\R\L E R L G RjK H Tl k]LjP V H RjZ\RjK H R]^ H PSR K E Z XlR Z V c G Z VYP T?R6^ E J`^ RjK ^ RjZ PSR]^T G kjXlRjK H ^3Tl J?KdRjK ^ RjZ PSR T?R M N E=G K H ^ e J G NgRjJeYRjK H jH X]R ETH RjKgJ`^ N VYX G K H RjXl^ R]L H G E KBVTYR]LPSR]^ kjPSkjZ\RjK H ^nT?R A o

- 7/.l /9 A CFE=GIH A J!K'R LjP V=^]^ R T?Rn^ E J`^ RjK ^ RjZ PSR]^6T?R Rd VTYR]L A 6= ∅o f V T G Z\RjK ^ G E K t E JBT G Z\RjK ^ G E K T?RWVjNQK G mgRjX E K'RjK =G ^ u VA T?R A R]^ H T?k)K G R N VYX

VA = supM ∈ N : S(A,M) = 2M,Gfo R o VA R]^ H PSR NQP J`^wrYXVYKQTBRjK HfG RjX M H RjP e J G P Rc G ^ H R6J?K RjK ^ RjZ PSR T?R M N E=G K H ^ e J G NgRjJ H jH XlRk]LjP V H kUN VYX A o

- 7/.l /9 A CFE=GIH G J!K&R LjP V=^]^ R T?R # E K'L H G E K ^T?R Rd NQX]RjKQVYK H ^ R]^s=VYPSRjJ!Xl^ T`VYK ^ RoHp K

K E=H R G+ PSR_^ E J^ RjK ^ RjZ PSR3T?R Rd+1 T?k)K G N VYXG+ =

(z, t) ∈ Rd × R; t ≤ g(z), g ∈ G

^ E=GIH P RjK ^ RjZ PSRT?R]^_^ E J`^rYX VjNdmgR]^T?R4# E K'L H G E K ^T?R G o

465879;:=<?>@BA CFE=GIH G J!K'R LjP V=^]^ R T?R # E K'L HfG E K ^ g : Rd → [0, B] VTYR]L 2 ≤ VG+ < ∞ a ^ E=GIHp ≥ 1 a ^ E=GIH ν J?K'RZ\R]^jJ?XlRnT!RiNQX ET V G P GIH k]^%^jJ?X Rd aiR H ^ E=GIH 0 < ε < B

4

o P E Xl^

N (ε,G, ‖ · ‖Lp(ν)) ≤ 3

(2eBp

εplog

3eBp

εp

)VG+

.

465879;:=<?>@BA CFE=GIH G J!K R]^ N V?L]RqYR]L HE X G RjP T?R\T G Z\RjK ^ G E K r T?R$# E K'L HfG E K ^\T?R Rd → R R H^ E=GIH

A =z : g(z) ≥ 0 : g ∈ G

.

P E X]^VA ≤ r.

^@LL

Page 213: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

. >%$ +! #!? 7% $ 2"+ 0 #"*$' 3k2( 0 / #` ' ! 2 " ' 2" 7 0 # $ 40D0 ,#

7% n2"*$Q! 2" 7* 0 , 0 #!? 2">@?"!1!ABACDBK^e?DF GrqAC A"b"BFON\[F

MN` P[GWVWBP[U]VYATC"B VYC"("`!FOC"(dP[CEU]FOB("? +ZA?1!"GWF

(X,Y )C"ATU`ZFOB

(Xm, Ym)1≤m≤MmhC C"ATUF

FM?"C FZB !dP+ZF [F+U]AN(VWFOG=(DF cATC"+OUVWAC"Bu^E?"V0!FZ?DU3("`!FOC"("NF("F

(X1 · · · , XM)mZhC(D`<]dC"VMU

mM

WdGWPjcATC"+OUVWAC ("FHN(`F`TNFZB(BVWATCcFZB(UVWRS`ZFIW +OARSR Fj¡

mM(·) = arg infψ∈FM

1

M

M∑

m=1

|ψ(Xm) − Ym|2.hCcVWC\U]N(A ("?"VYUGQP C"ATU]P[U]VYAC

‖ψ‖2M =

1

M

M∑

m=1

|ψ(Xm)|2..GYANB<W"AC PvGQP U]aD`ZAN(XZRSFwB?"Vk[[PTC\U8&Ua"`ZATNXZRSFqLTL1mYL.!"P$`FqL z | (dPTC"B0# x\^oSVU¡465879;:=<?>@BA C J N`N E ^ E K ^ e JgR

σ2 = supx∈Rd

V ar(Y |X = x) <∞.

CFE=GIHKM P V T G Z\RjK ^ G E K T?R FM a VYP E Xl^

E

(

‖mM −m‖2M

∣∣∣∣X1, . . . , XM

)

≤ σ2KM

M+ min

ψ∈FM

‖ψ −m‖2M .

C;G T?RUNQP J`^]a FM R]^ H G KQT?k NgRjKQT`VYK H T?R]^3Xlk VYP G ^V H G E K ^ (Xm, Ym)1≤m≤M a E KdVE

(

‖mM −m‖2M

)

≤ σ2KM

M+ min

ψ∈FM

E|ψ(X) −m(X)|2.6 FJU]a"`OANXORSFqLL1m_^/!dP$`FvL z /(dPTCDB=# x\^yS C"AT?"B-(DAC"C"F465879;:=<?>@BA CFE=GIH G J!K'R LjP V=^]^ R T?Ri# E K'L H G E K ^ g : Rd → R

jE XK'k]R*N VYX B R H ^ E=GIH ε > 0o

P E X]^P(∃g ∈ G : ‖g‖ − 2‖g‖M > ε

)≤ 3EN2(

√2

24ε,G, X2M

1 ) exp(− Mε2

288B2)

E&%X2M

1 = (X1, · · · , XM , XM+1, · · · , X2M ) VTYR]L (XM+m)1≤m≤M J?K'RUL E N G R GfoSGfo T_T?R (Xm)1≤m≤Mo

%$ $ 1"*, +!,<7* g2U k7%"465879;:=<?>@BA e CFE=G RjK H X1, · · · , XM T?R]^*=VYX G V PSR]^ VYPSk V HE=G XlR]^Xlk]RjPIPSR]^ G KQT?k NgRjKQT`VYK H R]^ oCFE=G RjK Ha1, b1, · · · , aM , bM T?R]^ Xlk]RjP ^ R H E K ^jJON`N E ^ R e JgR Xm ∈ [am, bm] VTYR]L NQX ET V G P GIH k 1 N E J?X1 ≤ m ≤M

o P E X]^]a ∀ε > 0

P

(∣∣∣∣

1

M

M∑

m=1

(Xm − E(Xm))

∣∣∣∣> ε

)

≤ 2 exp

(

− 2Mε2

1M

∑Mm=1 |bm − am|2

)

^@L ^

Page 214: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

.nm <!m . . 6 >D . K 6;6 #) ) )

. !"GY?"BVYFZ?"N(B NF!"NVYBFOBFW ATC3("AVMU:!"N(`+OVWBFON^E?"FOGWGWFHB(AGW?DUVWACcATC3+,aDAVWB(VYU-("? !DNA"b"GYXZRSF0("FkRSAVWC"(DNFZB+ P[NN`OBE¡

infα

1

M

M∑

m=1

|Ym − α.p(Xm)|2

A10p(·) FZB(UH?DC"FbdPTB(F(DFcAC"+U]VWATC"B("`!FZC"("PTCEUk`X[FZC\U]?"FOGWGYFZRSFZC\U0("FOB0(DAC"C"`OFZB

(Xm)1≤m≤MFU

("F + PTN ("VWCdP[GK

&AC6B(?"!"!AB(FK ≤ M

Ujm4C8A?@U]NFIW AC6B?"!"!ATBF:B(A?-[FZC\UA^E?"FB]PTCDB>!FZN(UF ("F``OC"`ZNPTGWVMU]` B∗B

M= Id

A10BFZB(U$GQP R P[U]N(V+ZF5(DFwU]PTVWGYGYF

M ×K("AC\U$GWFOB GYV2`CDFZB B(AC\U

p(Xm)∗m

G Cxqtn P !dPTB8(DF !DNA"b"GYXZRSF*("PTC"B GWF*+ZPTB A10 GQPcRsP&U]NV+ZFB∗B

FOB(U VWCR[FZNB(V2b"GYF*+ZPTN VYG FXp@VWB U]Fs?"C"F?"C"V1^E?"F B(AGY?DU]VYAC PT?>!"NA$b"GWXOR F ("FR AVYC"("N(FZB'+ P[NN`OB4FU4?"C !DNA +Z`("`@("F > +na"R V(DU !FONRSFOUUN]Pb"VWFOCW!dPTN ?"CDF U]N]PTCDB\cAN(RsP&U]VYACHGWVYC"` PTVYN(F (DFiGQPbdPTB(F

p(·) (qAN(Ua"AC"ATNR PTGWVMBFONlGWFOB +ZAGYAC"C"FOB'("FiGWP$RsP[UNV+ZFBm

- PTN=+ZATCEU]N(F$W (dP[C"B GWF*+ZPTB A10 VWG FXp@VYB(U]F ("FOB/+OAGWVYC"` P[NVMU]`OBHFZC\U]NFsGYFZB=+ZAGYAC"C"FOB8("FB

W VYG +ZACR[@VYFZC\U("F +,aDAVWB(VWN ?"C"F:R PTC"VYXZN(F ("FN`OBA?"(DNF:GWF !"N(A"b"GYXZRSF ("FR ATVWC"("N(FZB/+ZPTNN(`ZB &V mFYm'("F +,aDAVWB(VWN ?"C"FBATGW?DUVWAC !dPTN(UV+O?"GWVYXONFyU^e?DVBAVMU@+OAR!dP[U]V1b"GYFPy[F+wGWFcPTVMU ("FJB(?"!"!AB(FZN B∗B

M= Id

BPTC"B@!FZN U]F("F``OC"`ZNPTGWVMU]` m $C"F.!ABBV1b"VYGWVMU`FZB U GWPvB(?"V1[[P[CEU]F mhC FXv F+OU]?DF ?DC"F (D`+OAR*!ATBVYUVWAC6FZC [[PTGYFZ?"N(BSBVYC:`?"GYVWXONFZB& >6aW[AVYN#Q 6 % TS U/("F B√

M

^E?"VBFq`+ONVMU B√

M= UΣV ∗ P[F+

U?DC"FR P[U]N(V+ZF$AN U]a"A"`TACdPTGYF/&

U ∗U = IdU (DFUnP[VWGYGYF

M ×MWΣ?"CDF

R P[U]N(V+OF(DFHGWPjcATNRSF

Σ =

σ1 0 · · · 0

0 σ2 · · · 0mmm0

m m m0

0 . . . 0 σK0 . . . . . . 0mmm

. . . . . .

mmm0 . . . . . . 0

A10σ1 ≥ σ2 ≥ · · · σK ≥ 0

BATCEU GWFOB[[P[GWFO?"NB B(VWC:`?DGWVYXZNFOB ("F B√M

FUV?"C"FwR P[UNV+ZFAN(Ua"A"`ATCdPTGYF

("FsUnPTVYGWGMFK × K

m4>@V U]A?@U]FZB GWFOBa[[PTGYFZ?"N(B BVWC-`?"GWVYXZN(FZB B(ACEUvB(UNV+OU]FOR FOCEU/!ABVMU]Vk[FZBvPTGYANBkAC6BFNFU]N(A?-[F (dPTC"B GYF+ZP("NF (q?"C"F BATGW?DUVWAC8?"C"V2^E?"F mi>@V FZGYGWFZBqBAC\UsC"ATC Ce?DGWGYFZB (?"B\^E?q ?"C NPTC:`r < K

!D?"VWBσr+1 = · · · = σK = 0

WPTGYANBsAC +na"AVWB(VYU ("F1NFZR!"GWP+ZFONB

!dPTN[u1, · · · , ur]GWFOB

r!"NFORSVWXZN(FZB +OAGWATC"C"FOB(DF

Um- P[N !"NA !DNVW`U]` ("F GQP >6 &d[ATVWN #Q 6 % TSVUXW AC BPTVYU^E?"F

+ZFOBr

!"N(FZRSVWXONFZB-+ZATGWACDC"FZBcATNRSFZCEU$?"C"FCbdPTB(F ATN(U]aDAC"AN(RsP[GWF("? BAT?"BM_.FZB !"P+ZFC[F+OUANVYFZG?("FRMFZC-`FZC"(DN`!"PTNvGWFOB>+ZAGYAC"C"FOB (DF

Bm44C A?DUNFIW4VYGiFZB U PTGWATNBacP+OVWGYF ("Fs[`ZN(V1]dFON^E?"F:GQP1B(AGY?DU]VYAC

("? !"NA$b"GWXOR F5("FJRSAVWC"(DNFZB +ZPTNN(`ZB ?DC"FQcAVYB +ZFUU]FJUN]PTCDB\cAN(RsP[UVWATC F<v F+U]?"`OFHFZB U-("AC"CD`ZF.!dPTNGQPBATGW?DUVWAC ("FwC"ATNRSFJFZ?"+OGWV(DVWFOC"C"FJRSVWC"VYRSPTGYF m4C +ZAC1+ZGW?DBVWATCWATC P*("AC1+ B !`+ZV1]"`j^E?"FOGWGWFHBAGY?DU]VYAC P[? !"N(A"b"GYXZRSF=(DF R ATVWC"("N(FZB-+ P[NN`OB$+,a"ATVWBVYN

^@L

Page 215: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

. >

FOUC^E?"F GrqAC !A?-[[P[VYUEb"VWFOC B(?"!"!AB(FZNa^E?"F GWFOB0+ZATGWAC"CDFZB0("F B√M

B(ACEUkAN(Ua"AC"ATNR PTGYFZB<W ^E?"VYU(U]F NFOR*!DGQP+OFZN

K!dP[N

r < Km

^@L|

Page 216: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

F

# .K % 6 m.C"(DFZNB(FZC PTC"( K3miN(ATU]a"FON(U]ATC K$P[U +ZGWVkv F m pDP+U F p@ATU]V+ZBm !G ^ % ¡ z < ez % L % % m#tPTG % mPTGYG1nYm3.:!"!"NAp@VWR P[U]VYATC1B +na"FOR FjcANBAGY?DU]VYAC"BA$c >m K*V?L + VYXT\^ HE

Lnm V=^ H G L T G RjXlRjK H G VYP*R e J V H G E K ^ t VYX G ^]a= nu aiT GIH R T o P6VYX E J GVYKQT f o V o P G V a E P J!Z\R E # GIH Z VYK ! R]^ o E=H R]^ V H m o C RjX o a N Vr?R]^ A o f8E KerYZ V1K!a h VYXP E + L % % ?m#tPTN]x^ m m"P[NGWA 3m?. (DV1v ?"BVYACkR A (DFZGANlFZGYF+OUNV+ZVYU.!DNV+OFZBpm dV H mgRjZ V HfG LlVYP G KQVYK'L]R L ^s¡ ^ !\#"D^ % !$d^TxTx\^!m

#t wx\^ myPTGYG%& m_Um P 'dPTGWGYFZN(AF%m dFZN(CdPTC"("F& PTC1( m 4G wPTNA?DV m[K$F,dF+OU]F(a >)(B*-)(IBP[C"(,+[PTN(VQP[UVWATCdPTG VWCDF*-E?dPTGYVYU]VMFZBpm $! ( ! R N E X H A ^[xx\^!m#t xDL. - miNVWPTC"(*m FZG%ATC P[C"(0/ m `OR VYCFm ATC"B5TFON1.U !F Ua"FZAN(FZR2ANf > 4Bm PSR]L H X E K o E Z Z o X ET V o3 g¡YL4"L65d^TxxDLYm

#t87x mImwiFZC\U]a 6 m eGWPTC"( Km97 PT?:F PTC"( %m m $VWFOGWB(FZCFm<;C P[N='"VMU]N]P :F4>NFOF0!"NV?+ZVYC:HA&@ATN PTN (*+OACEUN]P+U]B*VWCSFZC"FON=: SR PTN5TFU]Bm N`NQP G R T dV H mgRjZ V H G L VYP$ G KQVYK'L]R Lx &A5B¡\^ <#" $"^Txx m

#tiFZN % < mbm FZN=:TRsP[CFmC;5!DUVWAC !"NV+ZVWC$: VMU]a ("V%D FZN(FZC\U]VQP[GVWCEUFZN(FZB(UHN]P[U]FOBm:! R o E #E G KQVYK'L G VYP C;H J T G R]^ F!"& ^GBIHJ5E <6"1<Txx$ L % % <?m#JK x&5& K%mGK A?"+nadPTN (@ mT8LTFOGQP[C"(PTC"( m A?ZV mM;C Ua"F PTGWGYVWP6+eVWC P!"!"N(A\P+na U]A ACEUFPTNGYAP2!"!"NA6N@VWR P[U]VYATCHA +OAC"("VMU]VYACdPTGF4N !F+OUnP&U]VWATC"Bm G KQVYK'L]R C;H E Lnm V=^ HfG L]^ O!PHQ5 <6"@LG^TxTx&5Wm

#JK % <O m mRKiPTNGYA PTC"(TS6m $?dPTGQP[N(U mwf R]L H J?XlR]^ E K NQX E V G P GIH H mgR E X VYKQT\^ H V HfG ^ H G L]^]aL E PSR Tl k H kBT?RhNQX ET V G P GIH k]^ T?R C V G K H > P E J?X U+AGY?"R F ^ < A& f R]L H J!XlR E=H R]^ G KdV H mgRjZ V H G L]^=& m > !"N(VWC:TFZN*L %% <!m#JKi- x V mKPTGWGW P[C"(hm1- P&:XOBm4N(NAN PTCdPTGWeBVWBA& Ua"F A !DUVWR PTG-E?dPTC\U]V ZP[U]VYACqPTG3:TANVMUa"RAN AG'"B UnP+OGWF!"NA&'"GWFOR BpmiC;H E Lnm V=^ H G L X E L]R]^]^ o N`NQP o3L x & L#BIHML4"5\x^TxTx m#JKi--*x < V mKPTGWGWG hm=- P :XZB.TPTC1(X/ m -4NVYCEU]FOR*!DBm?.Y-e?"PTCEUV P&U]VWATCHUNFOFiRSFOUa"A (ZATN;!"NV+ZVYC:PTC"( aDF(:TVWC: R ?"GMU]V("VYRqFZC"B(VWATCdPTG. R FONV+ZPTC A!DU]VYAC"Bpm V H mgRjZ V HfG LlVYP[ G KQVYK'L]R L <"& LOB4d^TxxA<?m

Page 217: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

#JK >" m KiGWP+ L PTC"( m > +na"AGYFZBm aDF !"N(V+ZVYC:cA&A !DUVWACDBkPTC"( +ZAN !AN]P[UF GYVQP&'DVYGYV U]VMFZBpm E J!XKQVYP E # E P GIH G L VYP L E K E Z> !DLGL % m#JK x&5& K%m@KiA?"+nadPTN ( P[C"( nm A?ZV m SVYB +ONFOUF UVWRSF P!"!"N(A#N@VWR P[UVYAC PTC"( ATCEU]F P[NGWABVYRv?DGQP[UVYATCHA 'dP2+ LPTN ( B(U]A +nadP[B(U]V+ ("VWD FONFZC\U]VWPTG\F*-E?dP[UVWACDBm;C;HE Lnm V=^ HfG L NQX E L]R]^]^ R]^VYKQT H mgR G XnVjN`NQP G L V HfG E K ^ LLTL HYL <#"D^Tx$d^Txx 5Wm#ia"F % O S6m ia"F4+[PTC"+OFYm $?"RSFZNV+ P[G RSFOUa"A ("BAN'"P+ LiPTN (:B(UA +,a"PTB(UV+5("VWD FONFOCEU]VWPTGF*-E?dPOU]VYAC"Bm K KgJ?Z\RjX G L VYP Z\R H m E T=^ G K )KQVYK'L]R]aiT GIH R T f o oo !E r?RjX]^ VYKQT o VYP V "!"P&:FZB ^^#"D^ 5&5 L % % !m# 6 - x\^O Um GW`ZRSFZC\U6S6m 6 PTR 'FZN(UACP[C"( - m#-4NATU(U]FONm6.C PTC"PTG%eBVYBA&iP GWFZPTB(UJB=-E?dPTN(FZBNF.:NFOBBVYAC1R FU]a"A ( ATN%. R FONV+ZPTC A !DUVWAC !"NV+ZVWC$:Wm G KQVYK'L]R C;H E Lnm o3 & 5B HJ5G5 % "5E@L&^Txx\^?m

# >Dxx 6 m iGYF GWA <PTC1(%m >EU]NV+ LEGWPTC"(Fm K'RjX rT?RjX G =V HfG YR]^ o X G L G Ker VYKQT !UG ^ Z VYKQVTr?RjZ\RjK H m 6 P+ZVYRsP - ?$'"GWV+ P[UVWATC"B* 6 AC"(DAC^[xxx m

#JS % ^TP S6m S?SF PTC1( 6 m !"B(UFZVYCFmh. BBFU=!"NV+ZVYC: VYUa B U]A +nadPTB U]V+ (DV%D FZN(FZCEUVQPTG ?@U]VWGYVMU1m! R G R+ E #< G KQVYK'L G VYP C;H J T G R]^ < HJ5"LL4"5 $ L % % ^!m#JS % ^ '$ S6mIS? F PTC"( 6 m !"B(UFZVYCFm >eUA +,a"PTB(UV+ ("V%D FZN(FZC\U]VQP[Gw?DU]VYGWVMUYm L E K E Z\R H X G L V x & ^GBIHA<O" % 5 L % % ^!m

#JSFZC % ! >&mSFZC$:Wm >eUA +,adP[B(U]V+ R A (DFZGWBA&*FOC"FZN :& +ZATR RSA ("VMU!"NV+ZFZBJPTC"(1U]a"FOVWNP!"!"GYV+ZPU]VYAC"B H"R FZPTC N(F.+FONBVYAC VYUa(?"R!"B PTC"(cB !DV3LTFZBpm XlR NQX G K H L % % ! m#JS x&5& mS FZGQP[N?"FvPTC"( >&m FZC"A OV mn. ATN PTN ( KiP+ L iPTN ( B(U]A +nadPTB U]V+qPTG%:AN(VMU]a"R ATN?dPTBV? GYVYC"FZPTN -8S4Bm! VjN`N E X H K H RjXK'R A=a ! a L E PSR E P H R]Lnm?K Gfe J;R [^TxTx&5Wm#JS B- % / mGSA?:TGQPTB./ m P$PTC"(h- m`-4NA[UU]FONm $?"R FONV+ PTGeR FU]a"A ("B AN AN PTN ('dP2+ LP[N (B(UA +,adP[B(U]V+*("VWD FONFZC\U]VWPTG4F.-e?"P[U]VYAC"Bmc KgKQVYP ^ E # N`N&P G RlT X ET V G P GIH & BXH % 5\xO"% G!L % % !m

#JS?"( % % Km mS?"("GYF.1m K G # E XjZ L]RjK H XV1P'P G Z GIH H mgR E XlRjZ6^ m P[RX'DNV($:FB(U]?"(DVWFOB VYC:P($+&PTC+ZF( RsP&U]a"FORsP[UV+ZB.L % % % m#JS !G! / m S?"!dP2+ZA#+[PPTC"( Km / m K%m FOU]Bpm . B eR!DU]ATUV+ 'FZadP6+eVWANlA&dB UnP[UVWB(UV+ZPTGEFOB(UVWR P[U]ATNB

PTC"( A&iA !DUVWR PTG BAGY?DU]VYAC"BJA&B U]A +nadPTB U]V+SA!DU]VYR V ZP[U]VMAC !"NAG'DGWFZRSBm mgR KgKQVYP ^E # C;H V H G ^ HfG L]^ L HYL <eL #"L < 5 % L % !&! m#8:TGQxA<O S6m" :GYA&D m ACEUF PTN(GWA PTG3:TANVMUa"RSB8AN A !DU]VYR PTGB(UA !"!"VYC:vPTC1(B UnP[U]VYB(UV+ZPTGGYF PTN C"VWC$:Wm QE VjN`NgR VYX G K KgKQVYP ^ E # N`NQP G R T X ET V G P GIH ^[xxA<!m# 3-

+ % nm G wP[NA?"V %m wP!A?"((VQPTC UmO- PTN ("A?N9&>&mO- FOC: PTC"( m %m?"FZC"FYmEK$F,dF+ U]F( BAGY?DU]VYAC"BA& '"P+ LiPTN ( >S)(IBHPTC1( NFOGQP[UF( A&'"B(U]P+ZGYF/!"NAG'DGWFZRSB AN -8S)(Bm KgK o X ET V o3d^ <"& ^&BIH_Tx\^#"@\ L % % !m# 3- % nm 4G wPTNAT?"V> >&mF- FOC:PTC"( m %m?"FOC"F mZKiP+ L iPTN ( B U]A +nadPTB U]V+8("V%D FZN(FZC\U]VQP[GF*-E?dP[UVWATC"B VWC dC"PTC"+ZF m V H mgRjZ V HfG L VYP9 G KQVYK'L]R d1& LOBIHML4"DeLGL %% !m

^@L

Page 218: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

# % < nms4G wPTNAT?"V P[C"( m%mJ?DFZC"FYm S eCdPTRSV+ !"N(AG:NPTRSR VMC: PTC"( !"NV+ZVYC:A& +ZAC\U]VYC:FZC\U +ZGWPTVWRSB VWC VYC"+ZATR*!"GYFOUF8R PTN=LTFU]Bm C o E K H X E P p N H G Z o3 & LOBIH_^ % " $ L % % <!m

# 6 % % 7nm GWGYRqFZNSPTC"( - m 6 FO?LTFON(U m ?dPTC\U]VWGYFca"F(:VYC:Wm G KQVYK'L]RBVYKQT C;HE Lnm V=^ HfG L]^ ,H_^ <@L "D^ 5 L % % % m# 4> % L* 7nm GYGYRSFZNPTC"( m&> +na FZV ZFZNpm7 F(:VWC$: A& +ZAC\U]VWC$:FZC\U?+ZGQP[VWRSB ?"C"("FONlVYC"+ZATR*!"GYFOUF

VWC$ATNR P[UVYATCFm C N`NQP G R T^ HE Lnm V=^ HfG L VYKQVYP O^ G ^ t f8E KQT E K a/u +ATGW?"RSF < A&C;HE Lnm V=^ HfG L]^ E K E rYX o31!dP&:FOB&! % "5DL65WmJAN (DACcPTC"(KiN(F P+na $F AN=LF L % % L1m

# x\^ 6 m AN R m Aa"GWFON*Q.nmJN4 P&LF P[C"(7m PTG%L;mw T G ^ H X G J H G E Kd #XlR]R H mgR E X E #3K E KYN VYXVYZ\R H X G L XlR rYXlR]^]^ G E K m > !"N(VWC:TFZN$>@FZN(VWFOB VWC >eU]P[U]VYB(U]V+ZB.^TxTx\^!m# 6 xx >&m JN]P P[C"(T7nm 6 ?"B +,a$:&1m E J?KQT`V H G E K ^ E # e JWVYK HfGMo V HfG E Kq# E X NQX ET V G P GIH T G ^ H X G J HfG E K ^ m 6 F+U]?"N(F$ATU]FOB VWC P[U]aDFZR P[U]V+ZB.L Tx$e> !"N(VWC:TFZN1>V FZN(GQP&:$KiFZN(GWVYCe^TxTxx m# 6 xA< Um wA&'FOU69/ m - m 6 FZRSAN*PTC"( m P[NVYCFm . NF.:NFOBBVYAC 'dPTBF( cACEUF4 iPTNGYAsRSF4

U]a"A ( AN['dP+ L iPTN (cB(U]A +nadP[B(U]V+.("VWD FONFOCEU]VWPTG F*-E?dP[UVWAC"Bpm QE VjN`NgR VYX G K KgKQVYP ^ E # N`NQP G R T X ET V G P GIH ^Txx <!m# x < Um JAG'FOUvP[C"(8K3m ?"CDABm >@FOC"BVMU]VW+@VMU PTCdPTGW@B(VWBk?"BVYC: U PTGYGYVWP6+eVYC+ZPTG+O?"GY?"BPTC"(8RsP[N(U]VYC:\P[GYFOBmb.:!"!"GYV+ZP[U]VYAC U]A B(U]A +nadPTB U]V+:A !@U]VWR PTG +ZAC\U]NATG m QE VjN`NgR VYX G KC E J?XKQVYP E K E K H X E P8VYKQT p N H G Z GMo V HfG E K d^Txx <!m# V 6 % hm/wAGY?'qPTC"( %m m&V*PTC 6 A\PTC'm dV H X G c L E Z%NQJ H V HfG E K ^ o 1X T R T oKiPTGYUVWRqAN(F& S H a"FE/TAaDC"B 7$A !LEVWCDB C"VW+&mg-4NFOBBmNN+@VYV> % 5 !Fm%L % % !m

# P7 xDL* .6m AadP&U]B? 7 V%:\P m F P L P!"!"N(A#N@VWR P[UVYAC"Bpm . PTGYGWVWP6+eVYC +ZPTG+O?"GW?DB P!"!DNA\P+naFmdV H m o E Z%N o3dTx & ^B HML A<#"L ^d^[xxDL1m# P7*- x\^ .6m AadP&U]B?7 V%:\PSP[C"(1K3mQ- FOU(U]FZN(BBATCFm V PTNVWPTC"+ZFwNF(D?"+OUVWAC R FU]a"A (DB AN BVWR ?GQP[UVWATC:A& ("FZCDBVYUVWFOB ATC VYFZC"FON B !dP+OFYm_C E J!XKQVYP E K J?Z\RjX G L VYP KQVYP ^ G ^ 5\x HQ5DL4"5A<Tx$d^TxTx\^!m

# k>!G! mHP[N]P[U PTB1P[C"( >Qm_UmJ>@aDNF.+F m X E + K G VYKZ E=HfG E K VYKQT ^ H E Lnm V=^ H G L LlVYPSLjJ?P J`^ m> !"N(VWC:FON1>V FZN(GQP&: L % !G! m# k> % ! mHP[N]P[U PTBwP[C"( >&m_Uml>ea"NF4+FYm R H m E T=^ E # V H mgRjZ V HfG L VYP G KQVYK'L]R m > !DNVWC$:FZN.L % % ! m# 6 >@xDL* m .6m 6 AC$:B(U]P DP[C"(qm_>&m\> +na P[N(U m V*PTGW?DVWC:%. R FONV+ PTC A !@U]VWATC"B ' BVYR ?"GQP&U]VMAC H

P BVYR*!"GYFHGYF P[B(U .B=-E?dPTN(FZBP!"!DNA\P2+,aFmJ! R G R+ E # G KQVYK'L G VYP C;H J T G R]^ lL65 HMLL #"L65!$^Txx@L1m# 6 >@x\^ / m 6 ?1+ZVWP PTC"( Um >Qm > +na P[N(U mw4GWF+U]N(V+ZVMU !"N(V+ZFOBJP[C"( !A iFZN$("FZN(V%+[P[UV%+FOB H U+eV?

("FZC1+ZFTNATR U]aDF C"AN ("V+ !A iFON:F N +,adP[C:F m ! R G R+ E #c6RjX G =V HfG YR]^ ! R]^ R VYXlLnm <"& L#BIH <6"1<Tx$d^Txx\^?m# 6 % L. m 6 F(DA?NqPTC"( m PTGQP :N]PTC1(Fm X ET V G P GIH G K VYKQV?Lnm C N V?L]R]^ m> !"NVYC:FON*DL %% L1m

^@L

Page 219: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

# 6 xDL. K%m 6 P!F4eNF*PTC"(HUm FZR PTR miAR!FOUVYUV%+F cACEUF PTN(GWAR FU]a"A ("BATNUa"F !"NV+ZVYC:A& PTBVWPTC:A !DUVWAC"Bpm E J!XKQVYP E # E Z%NQJ H V HfG E KQVYP9 G KQVYK'L]R <"& L#BIH % "1< % ^TxTxDL1m#B-*> x\^ / m P$ - m -4NATUUFZN.[/ m>DPTC PTN UYC PTC"( >&m AN(NFOBm $?"R FONV+ PTG$RSFU]a"A ( ATN'dP+ LPTN ( B U]A +nadPTB(UV+.("VWD FONFZC\U]VWPTG F*-E?dP[UVWACDBm KgK o N`NQP o X ET V o3L ^"& LOBIHx\^#"DL ^Txx\^?m

#B- % 5& / m PU- mU-4N(ATUUFZN*PTC"( / m m AC: m >@AGW+eVWC: AN PTN ('dP2+ LP[N ( B U]A +nadPTB U]V+("V%D FZN(FZC\U]VWPTGeF*-E?dP[UVWAC"B F4N !"GWV+ZVMU]GW P A?"N B(UF!qB +,a"FORSFYm X ET V o mgR E X ! RjP V H R T G RjP T=^ % ! & BIH % " A< % L % % 5 m#xA< / m PP[C"( / m a"PTC:WmDK F!"N(FZB(FZCEU]P[U]VYAC"BPTC"( N(F*:T?"GQPTN(VYUVYFZB ANlBAGY?DU]VYAC"B U]A<K >$S4B VYUa NF,dF+U]VWATC"Bm QE VjN`NgR VYX G K C;H E Lnm V ^ H G L X E L]R]^]^ R]^\VYKQT H mgR G X\VjN`NQP G L V H G E K ^ ^Txx <!m

# ?dP % < S6m ?"PTGQPTN U m mgR dVYPIP G VT G K L VYPSLjJ?P J^ VYKQT X]RjP V H R T H E N G L]^ m6-4NAG'"P&'"VYGWVMU P[C"( VYUB.$!"!DGWV+ P[UVWATC"Bmd> !DNVWC$:FZN V FONGQP : $F ATN=LF L % % <!m# - PTN % ! Um8- P[N ("A?N'm KP+ LPTN ( B U]A +nadPTB(UV+>("VWD FONFOCEU]VWPTG F*-E?dP[UVWACDBHP[C"(T+@VYB +OABVMU BAGY?

U]VYAC"BA& B eB(UFZRSBwA& BFOR VYGWVYCDF PTN-!dPTN]P ' ATGWV+ PTC"( FOGWGYV!@U]V+n- S BJA&*B(F+OAC"( AN ("FONm K C;HE Lnm V=^ HfG L KQVYP ^ G ^hVYKQT ! RjP V H R T QE N G L]^ mgR R G P EhE X ^nm E Nga@ a f o R]LjXlRjJ^ R# E KQT=a o ' RjX T?R]aM o p ^ RjKQT`VYP a o C&o ^ H J?K'RjPQR T=^ o aM G X mYJ^ RjX !dP&:FOB % "L ^ L %% ! m

# - K>DxDL* mi- A[UU]FONB. / m2- m[KiA?"+nadPT?1(PTC"( Snm >@FOB(U]A#+eV+Ym 7$F($:F( ACEUF4 PTNGYA H GWA +[PTN(VQPTC1+ZF (DFZNVW+[P[UV%+F !"N(V+ZVYC: VYUagAG' (F+OU]VW+F !"NA&'dP&'"VYGWVMUVYFOBm m ^ o ^&! % & 5BIH <eL #"1<^<d^TxTxDL1m

# - a"Pxx 7nmU-4adP[R mS eCdP[R V+Lp aDF(:VYC:6VYC (DVWB +ONFU]F1U]VWRSF ?"C"("FON +ZATC"F +OAC"B U]N]P[VWCEUBm

C o E K H X E P p N H G Z o3 G! & GBIHA<#" &!\^d^[xxx m# - AG ! 5& S6mY- AGWGWPTN ('m E KdYRjX r?RjK'L]R E # ^ HE Lnm V=^ HfG L NQX E L]R]^]^ R]^ mD> !"N(VWC:TFZN>@FZN(VWFOBVYCq>EUnP[UVWB(UV+OBm> !"N(VWC:FON1>V FZN(GQP&: $F AN LM L % !&5Wm# - - % x Um#- PTN ("A?N PTC"( >Qm#- FOC:Wmn.@(dP!DUF( BAGY?DU]VYAC1A& PC'dP2+ LP[N ( B(UA +,a"PTB(UV+>(DV%D F4

NFOCEU]VWPTGF.-E?dP[U]VYACFmiC ^ H RjZ6^ E K H X E P o f R HfH+o3 L65,H < <#" @LGL % % x m# - - % ^O Um - PTN ("A?N PTC"( >Qm - FOC:Wm KP2+ LP[N ( B U]A +nadPTB(UV+("V%D FZN(FZC\U]VQP[G F*-E?dP&U]VWATC"BsPTC1(-E?dPTBVYGWVYC"F P[N8!dPTN]P 'AGWV+ !dP[N(U]VWPTG ("VWD FONFZC\U]VWPTG F*-E?dP[UVWACDBm f R]L H J!XlR nE=H R]^ G K C aC NQX G Ker?RjX ;RjXP Vr L H_^TxxO"@^@L L % % ^!m# - N(V % % / m 6 m -4NV%:FZC\U m C1+ZAR!"GWFU]F RsPTN LTFOUB<H iATC+FON=:FOC"+ZFJA& A !DUVWAC"B +[P[GW?"FOB?"C1("FZNiU]aDFR VYC"VYR P[G RsP[N(U]VYC:\P[GYFR FZPTB?"N(FYm TT o N`NQP o X ET* DL HYLx <&!O"L x\ L % %% m# - N(A % x - mO- N(ATUUFZNmWC;H E Lnm V ^ H G L G K H R rYXV H G E K VYKQT T G RjX]RjK HfG VYP R e JWV HfG E K ^ oE K'R+ VjN`NQX E V?Lnm +AGY?"R Fw^@LwA&4 N`NQP G L V H G E K ^ E # V H mgRjZ V HfG L]^ m*> !DNVWC$:FZN V FONGQP :$KiFZN(GWVYC L % % x!m#K wxx Km K$AT?:F PTC"( nm 4G wPTNA?DV m%-4NV+ZVYC: +eVQP ?DU]VYGWVMU:R PN@VWRSV ZP[U]VMAC PTC1( FOCEU]N(A !1mdV H mgRjZ V H G L VYP9 G KQVYK'L]R Lx & ^GBIH_^ < % "D^d^Txxx!m

^@L#!

Page 220: Approximation par projections et simulations de Monte ... · Approximation par projections et simulations de Monte-Carlo des equations di erentielles stochastiques r etrogrades. Jean-Philippe

#K % m / m*K ?"C:&:\PTG("VWFONqP[C"( > +na iFOV OFZN m iAC+FZN :FZC"+OF A&A!DU]VYAC +[PTGY?"FZBS?"C1("FZNVWC"+OAR!"GWFU]FOC"FZB(Bm K C RjZ G KQVYX E K CgHE Lnm V=^ HfG L KQVYP O^ G ^]a R T=^H o E P H m VYJ^ RjK a o EZonoG VYKQT ol! J`^]^ E !"P&:FZB <#" G! 5 L % % m#K$AG:x\^ 6 m %m hm K$AG:TFZN(Bm AC\U]F PTNGYA +[P[GW?dP[UVWACcA&. R FONV+ZPTC A !@U]VWATC"Bm V H mgRjZ V HfG L VYP G KQVYK'L]R L^ H_^@L "D^&!$"^Txx\^?m#I> +,a % 5& mD> +naiFZV ZFONm8.$!"!DNA#N@VYRsP&U]VYC:wN]PTC"(DAR +[PTNVWP&'"GYFZB8'SB U]A +nadPTB U]V+$VWC\U]F*:TN]PTGYBm KdKQVYP ^ E #iNQX ET V G P GIH ^T^ HML <O"L <T < L % % 5 m#I> +,a % < m > +na FZV OFZNpm ;C U]a"F RSVWC"VYRSPTGsR PTN(UVWC:PTGMF R FZPTB?DNFgPTC"( Ua"F GWGYRqFZN > +na iFOV OFZN-("F+ZAR!ABVMU]VYACFm C;H E Lnm V=^ H G L KQVYP o N`NQP o3L & <GBIH <#"1< % % L % % <?m#I> +,a"xDL* m > +,aiFZV ZFONm . :?"V(DF( U]A?DNwU]a"NAT?:a -e?"P("N]P&U]V+qa"F(:TVWC: P!1!"NA\P2+,aDFZBm8p NH G E K NQX G L G Ker=a G K H RjX]R]^ H XV H R]^_VYKQT X G ^ Z VYKQVr?RjZ\RjK H a o E J G K G a o GIH VYK G L]a o J`^ G RjP V1tR T=^ o u 1!dP&:TFZB:<G!O" < 5"^TxxDL1m

# VJK % % / mm B(VYUBV3LEGYVWBSPTC"( K%m8V PTC K$A#1m ;5!DUVWR PTGB(U]A!"!"VWC$: A&RsPTN LTAO+ !"N(A +ZFOBBFOBTH7 VYG3'FON(UHB !dP+OF Ua"FZATN GlP!"!"N(A#N@VYRsP[UVYAC P[G3:AN(VMU]aDRSBwPTC"( PTC P!"!DGWV+ZP[UVWATC U]A!"NV?+ZVYC: a"V3:Ta ("VYRSFOC"BVYACdP[G "CdPTC"+OVQPTG (DFZNVW+[P[UV%+FOBm XVYK ^ o J H E Z o E K H X E P 5G5 & LxB HYL6!&5\xO"L6!A<@LG L % %% m

# V S)V % .6m m V*P[C SFON V*PPTN U PTC1( / m .6m FOGWGYC"FZNpm R V L E KdYRjX r?RjK'L]R VYKQT RjZ%N G X G L VYPNQX E L]R]^]^ R]^]a4+ GIH mBVjN`NQP G L V HfG E K ^ H E ^ H V HfG ^ H G L]^ m*> !"N(VWC:TFZN1>V FZN(GQP&:$ $F [ AN=LF L % % !m# adPxDL* / m adP[C:Wm >eAR F "C"F !"N(A !FZN U]VYFZB4A&M'dP+ LPTN ( B U]A +nadPTB U]V+@(DV%D [email protected]?"P[U]VYAC"Bm m\ H mgR]^ G ^]a J?X TYJgR K G YRjXl^ GIH ^[xxDL1m# adPx&5& / m adPTC: m1.8Ce?"RSFONV+ZPTGB +na"FOR F8ATN K >S4Bm mgR KgKQVYP ^ E # N`NQP G R T X ET V G P GIH L65 & LOBIHJ5 < % "5!G!$"^Txx&5Wm

^@L %