from particle filter to malliavin estimator with application parti… · malliavin esmator if...

35
From Par(cle Filters to Malliavin Filtering with Applica(on to Target Tracking Sérgio Pequito Phd Student Final project of Detection, Estimation and Filtering course July 7th, 2010

Upload: others

Post on 25-Jul-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

FromPar(cleFilterstoMalliavinFilteringwithApplica(ontoTarget

Tracking

SérgioPequito

PhdStudent

Final project of Detection, Estimation and Filtering course July 7th, 2010

Page 2: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Guidelines

■ Mo(va(on

■ MonteCarloMethods■ ImportanceSampling■ ControlVariates

■ Par(clefilters■ MalliavinEs(mator

■ Example

Page 3: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Mo(va(on‐Barelyrelevantsampling

Page 4: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MonteCarloMethods

Page 5: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Importancesampling

Importance sampling involves a change of probability measure. Instead of taking X from a distribution with pdf , we instead take it from a different distribution with pdf

where is the Radon-Nikodym derivative.

Page 6: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Importancesampling

We want the new variance to be smaller than the old variance

How to achieve this? By making small where is large, and making large when is small.

Small large relative to so more random samples in region where is large

Particularly important for rare event simulation where is zero almost everywhere

Page 7: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

ControlVariates

Suppose we want to approximate using a simple Monte Carlo average

If there is another payoff for which we know , can use to reduce error in

How? By defining a new estimator

Unbiased since

Page 8: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

ControlVariates

For a single sample

For a average of N samples

To minimize this, the optimum value for is

Page 9: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

ControlVariates

The resulting variance is

Where is the correlation between and .

The challenge is the choose a good which is well correlated with - the covariance, and hence the optimal , can be estimated from the data.

Page 10: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Remarks

Importance sampling – very useful for applications with rare events, but again needs to be fine-tuned for each application

Control variates – easy to implement and can be very effective But requires careful choice of control variate in each case

Overall, a tradeoff between simplicity and generality on one hand, and efficiency and programming effort on the other.

Page 11: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Par(clefilters

Page 12: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Bayes’Theorem

Page 13: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

NonlinearFiltering

■  Stateevolu(on■  discrete4mestochas4cmodel(Markovian)

■ Measurementequa(on

■ Modelpredic(ongivenpastmeasurements

Page 14: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

NonlinearFiltering

■  Predic(onupdateusingcurrentmeasurement

likelihood dynamic prior

posterior

Page 15: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

t

P(x)

x

tk tk+1

x actual state value measured state value state particle value state pdf (belief)

actual state trajectory estimated state trajectory particle propagation particle weight

Par(cleFiltering■  representstateasapdf

■  sample thestatepdfasa set of par(cles andassociatedweights

■  propagate par(clevalues according tomodel

■  update weights basedonmeasurement

Page 16: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

■ Predic(onstep:usethestateupdatemodel

■ Updatestep:withmeasurement,updatethepriorusingBayes’rule:

Par(cleFiltering

Page 17: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

■ A par(cle filter itera(vely approximates theposteriorpdfasaset:

where:

xkiisapointinthestatespacewkiisanimportanceweightassociatedwiththepoint

Par(cleFiltering

Page 18: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

■  Implementa(onsteps

■  proposex0iandpropagatepar4cles

■  computelikelihoodofmeasurementw.r.t.eachpar4cle

■  updatepar4cleweightsbasedonlikelihood

■  normalizeweights

Par(cleFiltering

Page 19: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Resampling

■ Par(cleweightsdegenerateover(me■ measureofdegeneracy:effec4vesamplesize

■  resamplewhenever■  newsetofpar4cleshavesamesta4s4calproper4es

use normalized weights

Page 20: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

PFFlowchartInitialize PF Parameters

Propose Initial Population , <x0,w0>

Propagate Particles using State Model , xk-1xk

Update Weights, wk-1wk Measuremen

t zk

Weights degenerated?

Resample

Yes

No

Page 21: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

Page 22: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

Particle Filter “Grid based method”

Page 23: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

Page 24: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

Page 25: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

If Hörmander hypothesis is satisfied then density exists and is smooth

Use Malliavin calculus to develop an expression for H(X,1) that can be Simulated. So we get a Monte Carlo for (with variance reduction)

Where are independent Euler approximations of .

For us then (pdf of X)

Page 26: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

The same simulated paths give good estimates for densities at any point. That is, one can compute the density over the whole real line with the same number of paths.

Let a localization function and a parameter and a “control variate”.

The density function

where

Page 27: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

The variance of is minimized for

where

and

Page 28: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

For the numerical implementation one have to discretize the following By “Euler” scheme

Moreover

Page 29: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

Page 30: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

Example

Page 31: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

MalliavinEs(mator

are W.G.N.

Page 32: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

N=1000 particles

Page 33: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

1000 runs!!!

N=20 particles

Brownian Paths: 50

Page 34: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

ConclusionandFurtherResearch

■ Different SMC method with variancereduc(on(noneedtocalibra(on)

■ Possibility toextend tononlinear caseandextendtohigherdimensions

Page 35: From Particle Filter to Malliavin Estimator with application Parti… · Malliavin Esmator If Hörmander hypothesis is satisfied then density exists and is smooth Use Malliavin calculus

References

35

B. Bouchard, I. Ekeland and N. Touzi (2002). On the Malliavin approach to Monte Carlo approximation of conditional expectations.

B. Bouchard and N. Touzi (2002), Discrete time approximation and Monte-Carlo simulation of Backward Stochastic Differential Equations

E. Fournié, J.M.Larsy, J. Lebouchoux, P.L. Lions, N. Touzi (1999). Applications of Malliavin calculus to Monte Carlo methods in Finance I

E. Fournié, J.M.Larsy, J. Lebouchoux, P.L. Lions (2001). Applications of Malliavin calculus to Monte Carlo methods in Finance II

D. Nualart (1995). The Malliavin Calculus and Related Topics.

Doucet, A.; De Freitas, N.; Gordon, N.J. (2001). Sequential Monte Carlo Methods in Practice

Doucet, A.; Johansen, A.M.; (2008). A tutorial on particle filtering and smoothing: fifteen years later

Arulampalam, M.S.; Maskell, S.; Gordon, N.; Clapp, T.; (2002). A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking