quantum algorithms for systems of linear equationsqml2018.umiacs.io/slides/somma.pdf · let’s...
TRANSCRIPT
![Page 1: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/1.jpg)
QuantumAlgorithmsforSystemsofLinearEquations
RolandoSommaTheoreticalDivisionLosAlamosNationalLaboratory
Jointworkwith
WorkshopattheIntersectionofMachineLearningandQuantumInformationUniversityofMaryland
September2018
RobinKothariMicrosoft
Yigit SubasiLosAlamos
AndrewChildsMaryland
DavideOrsucciVienna
![Page 2: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/2.jpg)
QuantumAlgorithmsforSystemsofLinearEquations
References:
• “Quantumlinearsystemsalgorithmwithexponentiallyimproveddependenceonprecision”,A.M.Childs,R.Kothari,andR.D.Somma,SIAMJ.Comp.46,1920(2017).
• “Quantumalgorithmsforlinearsystemsinspiredbyadiabaticquantumcomputing”,Y.Subasi,R.D.Somma,andD.Orsucci,arXiv:1805.10549(2018).
![Page 3: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/3.jpg)
Abriefhistoryofresultsinquantumcomputing
• SimulatingquantumsystemswasthemainmotivationbehindFeynman’sideaofaquantumcomputer(1982).
• Forexample,algorithmsforsimulatingthedynamicsofn spinsystemswithclassicalcomputershavecomplexitythatisexponentialinn.Quantumalgorithms,inprinciple,haveonlycomplexitythatispolynomialinn.
![Page 4: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/4.jpg)
Abriefhistoryofresultsinquantumcomputing
• SimulatingquantumsystemswasthemainmotivationbehindFeynman’sideaofaquantumcomputer(1982).
• Forexample,algorithmsforsimulatingthedynamicsofn spinsystemswithclassicalcomputershavecomplexitythatisexponentialinn.Quantumalgorithms,inprinciple,haveonlycomplexitythatispolynomialinn.
• PeterShordiscoversaquantumalgorithmforefficientfactorizationofintegerswithimportantapplicationstocryptography.Shor’salgorithmresultsinasuperpolynomial quantumspeedup(1994).Hisresultwasamainmotivationforthediscoveryofotherquantumalgorithms.
![Page 5: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/5.jpg)
Abriefhistoryofresultsinquantumcomputing
• SimulatingquantumsystemswasthemainmotivationbehindFeynman’sideaofaquantumcomputer(1982).
• Forexample,algorithmsforsimulatingthedynamicsofn spinsystemswithclassicalcomputershavecomplexitythatisexponentialinn.Quantumalgorithms,inprinciple,haveonlycomplexitythatispolynomialinn.
• PeterShordiscoversaquantumalgorithmforefficientfactorizationofintegerswithimportantapplicationstocryptography.Shor’salgorithmresultsinasuperpolynomial quantumspeedup(1994).Hisresultwasamainmotivationforthediscoveryofotherquantumalgorithms.
• L.Groverdiscoversaquantumalgorithmforunstructuredsearchresultinginapolynomial(quadratic)quantumspeedup(1997).AmainideainGrover’sresult(amplitudeamplification)hasbeenextensivelyusedinotherquantumalgorithmsforproblemssuchasoptimization,search,andmore.
![Page 6: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/6.jpg)
Abriefhistoryofresultsinquantumcomputing
• SimulatingquantumsystemswasthemainmotivationbehindFeynman’sideaofaquantumcomputer(1982).
• Forexample,algorithmsforsimulatingthedynamicsofn spinsystemswithclassicalcomputershavecomplexitythatisexponentialinn.Quantumalgorithms,inprinciple,haveonlycomplexitythatispolynomialinn.
• PeterShordescubre unalgoritmo cuántico parafactorizar enterosconaplicaciones importantes acybersecurity,resultando en unareducción exponencial delacomplejidad clásica (1994).Sibiencomputadoras cuánticas degrantamaño noexisten,este fue elprincipalresultado quedisparó investigación en algoritmoscuánticos
• L.Groverdiscoversaquantumalgorithmforunstructuredsearchresultinginapolynomial(quadratic)quantumspeedup(1997).AmainideainGrover’sresult(amplitudeamplification)hasbeenextensivelyusedinotherquantumalgorithmsforproblemssuchasoptimization,search,andmore.
Otherquantumalgorithmsforlinearalgebraproblems?
![Page 7: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/7.jpg)
Let’sconsidertheproblemofsolvingasystemoflinearequationsortherelatedproblemofinvertingamatrix:
LinearEquations:Animportantproblem
A.~x = ~
b ~x = A
�1~
b
![Page 8: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/8.jpg)
Let’sconsidertheproblemofsolvingasystemoflinearequationsortherelatedproblemofinvertingamatrix:
LinearEquations:Animportantproblem
A.~x = ~
b ~x = A
�1~
b
NxN N-dimensional
![Page 9: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/9.jpg)
Let’sconsidertheproblemofsolvingasystemoflinearequationsortherelatedproblemofinvertingamatrix:
LinearEquations:Animportantproblem
A.~x = ~
b ~x = A
�1~
b
• Thereisavarietyofclassicalalgorithmstosolvethisproblem.Nevertheless,evenwhenthematrixA andvectorƃ aresparse,thecomplexityof``exact’’classicalalgorithmsisatleastlinearinN.
![Page 10: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/10.jpg)
Let’sconsidertheproblemofsolvingasystemoflinearequationsortherelatedproblemofinvertingamatrix:
LinearEquations:Animportantproblem
A.~x = ~
b ~x = A
�1~
b
• Thereisavarietyofclassicalalgorithmstosolvethisproblem.Nevertheless,evenwhenthematrixA andvectorƃ aresparse,thecomplexityof``exact’’classicalalgorithmsisatleastlinearinN.
• Aresult[HHL08]:Quantumcomputerscanprepareaquantumstateproportionaltothesolutionofthesystemintimethatispolynomialintheconditionnumber,inverseofprecision,andthelogarithmofthedimension(undersomeassumptions).
![Page 11: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/11.jpg)
Let’sconsidertheproblemofsolvingasystemoflinearequationsortherelatedproblemofinvertingamatrix:
LinearEquations:Animportantproblem
A.~x = ~
b ~x = A
�1~
b
• Thereisavarietyofclassicalalgorithmstosolvethisproblem.Nevertheless,evenwhenthematrixA andvectorƃ aresparse,thecomplexityof``exact’’classicalalgorithmsisatleastlinearinN.
• Aresult[HHL08]:Quantumcomputerscanprepareaquantumstateproportionaltothesolutionofthesystemintimethatispolynomialintheconditionnumber,inverseofprecision,andthelogarithmofthedimension(undersomeassumptions).
• Note:Thisisasomewhatdifferentproblem(QLSP)andclassicalalgorithmsmaydobetterinthiscase.However,theQLSPisBQP-Complete.
![Page 12: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/12.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
![Page 13: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/13.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
![Page 14: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/14.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
![Page 15: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/15.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
![Page 16: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/16.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
Let CA(t, ✏) be the cost of simulating e�iAtwith precision ✏
![Page 17: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/17.jpg)
QuantumLinearSystemProblem(QLSP)
A.~x = ~
b
Assumptions
• A isHermitianofdimensionNxN• A iss-sparse• A isinvertibleanditsconditionnumberis𝜅<∞• ThespectralnormofA isboundedby1
Let CA(t, ✏) be the cost of simulating e�iAtwith precision ✏
![Page 18: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/18.jpg)
Hamiltoniansimulation
Note:RecentadvancesinHamiltoniansimulationresultedin
CA(t, ✏) = ˜O(tsTA log(t/✏))
• Complexityalmostlinearintheevolutiontime• Complexityispolylogarithmic intheinverseofaprecisionparameter
D.Berry,A.Childs,R.Cleve,R.Kothari,andRDS,PRL114,090502(2015)D.Berry,A.Childs,andR.Kothari,FOCS2015,792(2015)G.H.LowandI.Chuang,PRL118,010501(2017)
![Page 19: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/19.jpg)
QuantumLinearSystemProblem(QLSP)Someapplications:
• Inphysics,wherethegoalistocomputetheexpectationvalueoftheinverseofamatrix.Thisideawasusedin[1]forobtainingtheresistanceofanetwork.
• Instatmech,where,e.g.,estimatingthehittingtimeofaMarkovchainalsoreducestocomputingtheexpectationvalueoftheinverseofamatrix[2]
• InML,forsolvingproblemsrelatedtoleast-squaresestimation[3],byapplyingthepseudoinverse:
• Forsolvingcertainlineardifferentialequations[4]:
[1]G.Wang,arXiv:1311.1851(2013).[2]A.ChowdhuryandR.Somma,QIC17,0041(2017)[3]N.Wiebe,D.Braun,andS.Lloyd,PRL109,050505(2012).[4]D.Berry,A.Childs,A.Ostrander,andG.Wang,CMP356,1057(2017)
![Page 20: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/20.jpg)
QuantumLinearSystemProblem(QLSP)
Anote: Evenforthoseapplications,anumberofassumptionsmustbemadeinordertoobtainquantumspeedups.Theseassumptionsincludeefficientpreparationofcertainstates(ofexp manyamplitudes),nicescalingoftheconditionnumber,andsolvingcertainproblemslikecomputingexpectationvalues.Forthesereasons,shownquantumspeedupsaretypicallypolynomial.
![Page 21: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/21.jpg)
TheHHLAlgorithmfortheQLSP[5]
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
[HHL08]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
O [(Tb + CA(/✏, ✏/))]
![Page 22: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/22.jpg)
TheHHLAlgorithmfortheQLSP[5]
[HHL08]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
ConsideringthatmanyHamiltonianscanbesimulatedefficientlyonquantumcomputers,thecomplexitydependenceonthedimensionissmall(e.g.,logarithmic)
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
O [(Tb + CA(/✏, ✏/))]
![Page 23: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/23.jpg)
ImprovementsoftheHHLAlgorithm[6]
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)[6]A.Ambainis,STACS14,636(2012)
[HHL08]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
FurtherimprovementsbyAmbainis (VariableTimeAmplitudeAmplificationorVTAA):
[6]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
O⇥Tb + CA(/✏
3, ✏)⇤
ConsideringthatmanyHamiltonianscanbesimulatedefficientlyonquantumcomputers,thecomplexitydependenceonthedimensionissmall(e.g.,logarithmic)
O [(Tb + CA(/✏, ✏/))]
![Page 24: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/24.jpg)
ImprovementsoftheHHLAlgorithm[6]
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)[6]A.Ambainis,STACS14,636(2012)
[HHL08]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
FurtherimprovementsbyAmbainis (VariableTimeAmplitudeAmplificationorVTAA):
[6]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
O⇥Tb + CA(/✏
3, ✏)⇤
• NotethatthebestHamiltoniansimulationmethodshavequeryandgatecomplexitiesalmostlinearinevolutiontimeandlogarithmicinprecision
Almostlinearin𝜅!
ConsideringthatmanyHamiltonianscanbesimulatedefficientlyonquantumcomputers,thecomplexitydependenceonthedimensionissmall(e.g.,logarithmic)
O [(Tb + CA(/✏, ✏/))]
![Page 25: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/25.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
![Page 26: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/26.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
![Page 27: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/27.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)[6]A.Ambainis,STACS14,636(2012)
|bi !N�1X
j=0
cj |vjiI |�jiE
![Page 28: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/28.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
|bi !N�1X
j=0
cj |vjiI |�jiE
Thisregistercontainstheeigenvalueestimate(superposition):
• Itsufficestohavetheestimatewithrelativeprecision𝜖• Orderlog(𝜅/𝜖)ancillaryqubits
![Page 29: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/29.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
|bi !N�1X
j=0
cj |vjiI |�jiE
Then we implement the conditional rotation: |˜�jiE ! |˜�jiE
1
˜�j
|0iO +
s1� 1
2˜�2j
|1iO
!
![Page 30: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/30.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
|bi !N�1X
j=0
cj |vjiI |�jiE
Then we implement the conditional rotation: |˜�jiE ! |˜�jiE
1
˜�j
|0iO +
s1� 1
2˜�2j
|1iO
!
Undophaseestimation
![Page 31: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/31.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
|bi !N�1X
j=0
cj |vjiI |�jiE
Then we implement the conditional rotation: |˜�jiE ! |˜�jiE
1
˜�j
|0iO +
s1� 1
2˜�2j
|1iO
!
Undophaseestimation
Amplitude amplification for amplifying the amplitude of the |0iO state
![Page 32: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/32.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
Roughly,thescalingoftheHHLalgorithmcanbeanalyzedfromtheworstcase:
|bi = (1/)|v1/i+p
1� 1/2|v1i
![Page 33: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/33.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
Roughly,thescalingoftheHHLalgorithmcanbeanalyzedfromtheworstcase:
|bi = (1/)|v1/i+p
1� 1/2|v1i
Theactionof1/ A willroughlycreatetheequalsuperpositionstate,sobothareequallyimportant
![Page 34: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/34.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
Roughly,thescalingoftheHHLalgorithmcanbeanalyzedfromtheworstcase:
|bi = (1/)|v1/i+p
1� 1/2|v1i
Theactionof1/ A willroughlycreatetheequalsuperpositionstate,sobothareequallyimportant
ForthedesiredprecisionweneedtoevolvewithA fortimeoforder𝜅/𝜖
![Page 35: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/35.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
Roughly,thescalingoftheHHLalgorithmcanbeanalyzedfromtheworstcase:
|bi = (1/)|v1/i+p
1� 1/2|v1i
Theactionof1/ A willroughlycreatetheequalsuperpositionstate,sobothareequallyimportant
ForthedesiredprecisionweneedtoevolvewithA fortimeoforder𝜅/𝜖
Theactionof1/ (𝜅 A )onthestatereducesitsamplitudebyorder1/ 𝜅 andorder𝜅amplitudeamplificationroundsareneeded
![Page 36: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/36.jpg)
AquickviewoftheHHLalgorithmandVTAA
[5]Harrow,Hassidim,Lloyd,PRL103,150502(2009)
Roughly,thescalingoftheHHLalgorithmcanbeanalyzedfromtheworstcase:
|bi = (1/)|v1/i+p
1� 1/2|v1i
Theactionof1/ A willroughlycreatetheequalsuperpositionstate,sobothareequallyimportant
ForthedesiredprecisionweneedtoevolvewithA fortimeoforder𝜅/𝜖
Theactionof1/ (𝜅 A )onthestatereducesitsamplitudebyorder1/ 𝜅 andorder𝜅amplitudeamplificationroundsareneeded
FromhereweseethatweneedtoevolvewithA fortimethatis,atleast,order𝜅2/𝜖
![Page 37: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/37.jpg)
AquickviewoftheHHLalgorithmandVTAA
Howcanweimprovethistimecomplexitytosomethingthatisalmostlinearintheconditionnumber?
[6]A.Ambainis,STACS14,636(2012)
![Page 38: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/38.jpg)
AquickviewoftheHHLalgorithmandVTAA
Howcanweimprovethistimecomplexitytosomethingthatisalmostlinearintheconditionnumber?OneanswerisviaVariableTimeAmplitudeAmplification(VTAA)[6]
[6]A.Ambainis,STACS14,636(2012)
![Page 39: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/39.jpg)
AquickviewoftheHHLalgorithmandVTAA
[6]A.Ambainis,STACS14,636(2012)
Howcanweimprovethistimecomplexitytosomethingthatisalmostlinearintheconditionnumber?OneanswerisviaVariableTimeAmplitudeAmplification(VTAA)[6]Theroughideaisasfollows(again,consideringtheworstcase):
• Firstwedoabad-precisionphaseestimationtodistinguishlargefromsmalleigenvalues.ThismaybedoneevolvingwithA fortimeindependentof𝜅
• Thenweimplementaroughapproximationof1/𝜅 A toeigenstatesoflargeeigenvalue
• Weneedorder𝜅 amplitudeamplificationsteps• Weimplementanaccurateapproximationof1/ 𝜅 A toeigenstatesofsmall
eigenvalue• Amplitudeamplificationfororder1steps• UndophaseestimationorapplytheFouriertransform
|bi = (1/)|v1/i+p
1� 1/2|v1i
![Page 40: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/40.jpg)
AquickviewoftheHHLalgorithmandVTAA
[6]A.Ambainis,STACS14,636(2012)
Howcanweimprovethistimecomplexitytosomethingthatisalmostlinearintheconditionnumber?OneanswerisviaVariableTimeAmplitudeAmplification(VTAA)[6]Theroughideaisasfollows(again,consideringtheworstcase):
• Firstwedoabad-precisionphaseestimationtodistinguishlargefromsmalleigenvalues.ThismaybedoneevolvingwithA fortimeindependentof𝜅
• Thenweimplementaroughapproximationof1/𝜅 A toeigenstatesoflargeeigenvalue
• Weneedorder𝜅 amplitudeamplificationsteps• Weimplementanaccurateapproximationof1/ 𝜅 A toeigenstatesofsmall
eigenvalue• Amplitudeamplificationfororder1steps• UndophaseestimationorapplytheFouriertransform
ThecomplexityofVTAAintermsofprecisionisworsethanthatofHHL
|bi = (1/)|v1/i+p
1� 1/2|v1i
![Page 41: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/41.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[7]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
[7]A.Childs,R.Kothari,RDS,SIAMJ.Comp.46,1920(2017).
˜O [(Tb + CA( log(/✏, ✏/))]
![Page 42: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/42.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[7]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
• Thisresultsinanexponentialimprovementontheprecisionparameter
[7]A.Childs,R.Kothari,RDS,SIAMJ.Comp.46,1920(2017).
˜O [(Tb + CA( log(/✏, ✏/))]
![Page 43: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/43.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[7]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
• Thisresultsinanexponentialimprovementontheprecisionparameter• ItcanbeimprovedusingaversionofVTAAto:
[7]A.Childs,R.Kothari,RDS,SIAMJ.Comp.46,1920(2017).
[7]ThereexistsaquantumalgorithmthatsolvestheQLSPwithcomplexity
˜O [(Tb + CA( log(/✏, ✏/))]
˜O [Tb + CA( log(/✏, ✏))]
![Page 44: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/44.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• ThepreviousresultallowedustoproveapolynomialquantumspeedupforhittingtimeestimationintermsofthespectralgapofaMarkovchainandprecision(A.Chowdhury,R.D.Somma,QIC17,0041(2017)).
• Havingasmallcomplexitydependenceonprecisionisimportantfor,e.g.,computingexpectationvaluesofobservablesatthequantummetrologylimit.
Whytheseimprovementsareimportant?
![Page 45: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/45.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[8]ThereexistsaquantumalgorithmthatsolvestheQLSPbyevolvingwithHamiltoniansthatarelinearcombinationsof(productsof)A,theprojectorintheinitialstate,andPaulimatrices.Theoverallevolutiontimeis
[8]Y.Subasi,RDS,D.Orsucci,arXiv:1805.10549(2018).
O(/✏)
![Page 46: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/46.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[8]ThereexistsaquantumalgorithmthatsolvestheQLSPbyevolvingwithHamiltoniansthatarelinearcombinationsof(productsof)A,theprojectorintheinitialstate,andPaulimatrices.Theoverallevolutiontimeis
UsingHamiltoniansimulation,thistransferstocomplexity
[8]Y.Subasi,RDS,D.Orsucci,arXiv:1805.10549(2018).
O(/✏)
O(Tb/✏+ CA(/✏, ✏))
![Page 47: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/47.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• IwillpresenttwoquantumalgorithmsfortheQLSPthatimprovepreviousresultsindifferentways:
[8]ThereexistsaquantumalgorithmthatsolvestheQLSPbyevolvingwithHamiltoniansthatarelinearcombinationsof(productsof)A,theprojectorintheinitialstate,andPaulimatrices.Theoverallevolutiontimeis
UsingHamiltoniansimulation,thistransferstocomplexity
O(/✏)
O(Tb/✏+ CA(/✏, ✏))
• Themethodisverydifferentandbasedonadiabaticevolutions.Itdoesnotrequireofcomplicatedsubroutinessuchasphaseestimationandvariabletimeamplitudeamplification,thereforereducingthenumberofancillaryqubitssubstantially.
[8]Y.Subasi,RDS,D.Orsucci,arXiv:1805.10549(2018).
![Page 48: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/48.jpg)
Thistalk:twoquantumalgorithmsfortheQSLP
• PhaseestimationandVTAArequireseveralancillaryqubits(beyondthoseneededforHamiltoniansimulation)
• Withintwoweeksofpostingourresult,agroupimplementedouralgorithminNMR,claimingthatitisthelargestsimulatedinstancesofar(8x8)[9]
Whythisimprovementisimportant?
[9]J.Wen,et.al.,arXiv:1806.0329(2018)
![Page 49: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/49.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
![Page 50: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/50.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
![Page 51: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/51.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• 1/A isnotunitaryandweneedtofindaunitaryimplementationforit.Wethengothroughasequenceofapproximations:
![Page 52: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/52.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• 1/A isnotunitaryandweneedtofindaunitaryimplementationforit.Wethengothroughasequenceofapproximations:
wearegettingcloser:Linearcombinationofunitaries
![Page 53: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/53.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
![Page 54: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/54.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
�1.0 �0.5 0.5 1.0
�20
�10
10
20
![Page 55: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/55.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
�1.0 �0.5 0.5 1.0
�20
�10
10
20
![Page 56: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/56.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
�1.0 �0.5 0.5 1.0
�20
�10
10
20
Themaximum“evolutiontime”underA intheapproximationof1/A is
![Page 57: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/57.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• Sofarweapproximated1/A,withinthedesiredaccuracy,byafinitelinearcombinationofunitaries.EachunitarycorrespondstoevolvingwithA forcertaintime,andthemaxevolutiontimeisalmostlinearintheconditionnumber
![Page 58: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/58.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• Sofarweapproximated1/A,withinthedesiredaccuracy,byafinitelinearcombinationofunitaries.EachunitarycorrespondstoevolvingwithA forcertaintime,andthemaxevolutiontimeisalmostlinearintheconditionnumber
QLSP Hamiltoniansimulation
![Page 59: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/59.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
![Page 60: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/60.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
p�1|0i+
p�2|1i
| i Ui
V V †|0i
(�1U1 + �2U2)| i|0i+p�1�2(U1 � U2)| i|1i
![Page 61: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/61.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
p�1|0i+
p�2|1i
| i Ui
V V †|0i
(�1U1 + �2U2)| i|0i+p�1�2(U1 � U2)| i|1i
Correctpart Incorrectpart(orthogonal)
![Page 62: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/62.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
p�1|0i+
p�2|1i
| i Ui
V V †|0i
(�1U1 + �2U2)| i|0i+p�1�2(U1 � U2)| i|1i
Correctpart Incorrectpart(orthogonal)
This idea can be generalized to the case where the goal is to implement
M�1X
i=0
�iUi
![Page 63: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/63.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
p�1|0i+
p�2|1i
| i Ui
V V †|0i
(�1U1 + �2U2)| i|0i+p�1�2(U1 � U2)| i|1i
Correctpart Incorrectpart(orthogonal)
This idea can be generalized to the case where the goal is to implement
M�1X
i=0
�iUi
1
�(M�1X
i=0
�iUi)| i|0 . . . 0i+ |⇠?i
![Page 64: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/64.jpg)
Implementingalinearcombinationofunitaries
Suppose we want to implement the operator �1U1 + �2U2 to some state | iwhere �i � 0, �1 + �2 = 1, and Ui unitary
p�1|0i+
p�2|1i
| i Ui
V V †|0i
(�1U1 + �2U2)| i|0i+p�1�2(U1 � U2)| i|1i
Correctpart Incorrectpart(orthogonal)
This idea can be generalized to the case where the goal is to implement
M�1X
i=0
�iUi
1
�(M�1X
i=0
�iUi)| i|0 . . . 0i+ |⇠?i Amplitudeamplificationtoobtainthecorrectpart
![Page 65: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/65.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• WeusetheLCUapproachtoimplementtheFourierapproximationof1/A
![Page 66: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/66.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• WeusetheLCUapproachtoimplementtheFourierapproximationof1/A
• Note:WeassumethatthegatecomplexityoftheoperationVissmallwithrespecttoothercomplexities
• Addingupallthecoefficientsinthelinearcombinationofunitaries,weobtain
� = O()
![Page 67: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/67.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• WeusetheLCUapproachtoimplementtheFourierapproximationof1/A
• Note:WeassumethatthegatecomplexityoftheoperationVissmallwithrespecttoothercomplexities
• Addingupallthecoefficientsinthelinearcombinationofunitaries,weobtain
• Thisisalsothenumberofamplitudeamplificationrounds
� = O()
![Page 68: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/68.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• WeusetheLCUapproachtoimplementtheFourierapproximationof1/A
• Note:WeassumethatthegatecomplexityoftheoperationVissmallwithrespecttoothercomplexities
• Addingupallthecoefficientsinthelinearcombinationofunitaries,weobtain
• Thisisalsothenumberofamplitudeamplificationrounds
• Then,theoverallcomplexityofthisapproachis
� = O()
˜O [(Tb + CA( log(/✏, ✏/))]
![Page 69: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/69.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• WeusetheLCUapproachtoimplementtheFourierapproximationof1/A
• Note:WeassumethatthegatecomplexityoftheoperationVissmallwithrespecttoothercomplexities
• Addingupallthecoefficientsinthelinearcombinationofunitaries,weobtain
• Thisisalsothenumberofamplitudeamplificationrounds
• Then,theoverallcomplexityofthisapproachis
� = O()
Thisisalmostquadraticintheconditionnumber.ToimproveittoalmostlinearweuseaversionofVTAAthatdoesn’truinthelogarithmicscalinginprecision
˜O [(Tb + CA( log(/✏, ✏/))]
![Page 70: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/70.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• VTAAforHHLreliesheavilyonphaseestimation,bringingaprohibitivecomplexitydependenceonprecision
• Butinourcaseweonlyneedtodistinguishtheregionsfortheeigenvalueswithhighconfidence,sothescalinginprecisionislogarithmic
![Page 71: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/71.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• VTAAforHHLreliesheavilyonphaseestimation,bringingaprohibitivecomplexitydependenceonprecision
• Butinourcaseweonlyneedtodistinguishtheregionsfortheeigenvalueswithhighconfidence,sothescalinginprecisionislogarithmic
• ThefinalalgorithmisVTAAappliedtoanotheralgorithmthatisbuiltuponasequenceofsteps.
• Ateachstepwedothefollowing:i)Wedeterminetheregionoftheeigenvaluewithhighconfidence.ii)Weapply1/Awithinthenecessaryprecisionforthatregion(replacingtheconditionnumber)
![Page 72: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/72.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• VTAAforHHLreliesheavilyonphaseestimation,bringingaprohibitivecomplexitydependenceonprecision
• Butinourcaseweonlyneedtodistinguishtheregionsfortheeigenvalueswithhighconfidence,sothescalinginprecisionislogarithmic
• ThefinalalgorithmisVTAAappliedtoanotheralgorithmthatisbuiltuponasequenceofsteps.
• Ateachstepwedothefollowing:i)Wedeterminetheregionoftheeigenvaluewithhighconfidence.ii)Weapply1/Awithinthenecessaryprecisionforthatregion(replacingtheconditionnumber)
• Theoverallcomplexityofthisapproachis
˜O [Tb + CA( log(/✏, ✏))]
![Page 73: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/73.jpg)
Firstalgorithm:AFourierapproachforsolvingtheQSLP
• VTAAforHHLreliesheavilyonphaseestimation,bringingaprohibitivecomplexitydependenceonprecision
• Butinourcaseweonlyneedtodistinguishtheregionsfortheeigenvalueswithhighconfidence,sothescalinginprecisionislogarithmic
• ThefinalalgorithmisVTAAappliedtoanotheralgorithmthatisbuiltuponasequenceofsteps.
• Ateachstepwedothefollowing:i)Wedeterminetheregionoftheeigenvaluewithhighconfidence.ii)Weapply1/Awithinthenecessaryprecisionforthatregion(replacingtheconditionnumber)
• Theoverallcomplexityofthisapproachis
UsingthebestHamiltoniansimulationmethods,thisisalmostlinearintheconditionnumberandpolylog ininverseofprecision
˜O [Tb + CA( log(/✏, ✏))]
![Page 74: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/74.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
![Page 75: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/75.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• TheideahereistopreparetheeigenstateofaHamiltonianbypreparingasequencecontinuouslyrelatedeigenstatesofafamilyofHamiltonians
![Page 76: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/76.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• TheideahereistopreparetheeigenstateofaHamiltonianbypreparingasequencecontinuouslyrelatedeigenstatesofafamilyofHamiltonians
• Wewanttheeigenstatetobethedesiredquantumstate(aftertracingoutancillarysystems)
![Page 77: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/77.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• TheideahereistopreparetheeigenstateofaHamiltonianbypreparingasequencecontinuouslyrelatedeigenstatesofafamilyofHamiltonians
• Wewanttheeigenstatetobethedesiredquantumstate(aftertracingoutancillarysystems)
P
?b A.~x = P
?b~
b = 0
B
†B|xi = 0 , B = P
?b .A
![Page 78: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/78.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• TheideahereistopreparetheeigenstateofaHamiltonianbypreparingasequencecontinuouslyrelatedeigenstatesofafamilyofHamiltonians
• Wewanttheeigenstatetobethedesiredquantumstate(aftertracingoutancillarysystems)
Thefollowingpropertiescanbeproven:
• ThedesiredstateistheuniquegroundstateofH• Theeigenvaluegapisorder1/ 𝜅2
P
?b A.~x = P
?b~
b = 0
B
†B|xi = 0 , B = P
?b .A
H
![Page 79: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/79.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• TheideahereistopreparetheeigenstateofaHamiltonianbypreparingasequencecontinuouslyrelatedeigenstatesofafamilyofHamiltonians
• Wewanttheeigenstatetobethedesiredquantumstate(aftertracingoutancillarysystems)
Thefollowingpropertiescanbeproven:
• ThedesiredstateistheuniquegroundstateofH• Theeigenvaluegapisorder1/ 𝜅2
• WenowseekthefamilyofinterpolatingHamiltonians
P
?b A.~x = P
?b~
b = 0
B
†B|xi = 0 , B = P
?b .A
H
![Page 80: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/80.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
![Page 81: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/81.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
• Wedefinetheinterpolatingmatrix A(s) = (1� s)I + sA , 0 s 1
![Page 82: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/82.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
• Wedefinetheinterpolatingmatrix
• Similarly,wedefine
A(s) = (1� s)I + sA , 0 s 1
H(s) = B†(s)B(s) , B(s) = P?b A(s)
![Page 83: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/83.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
• Wedefinetheinterpolatingmatrix
• Similarly,wedefine
• Thisislikesolvinganincreasinglydifficultsystemoflinearequations!
A(s) = (1� s)I + sA , 0 s 1
H(s) = B†(s)B(s) , B(s) = P?b A(s)
![Page 84: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/84.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
• Wedefinetheinterpolatingmatrix
• Similarly,wedefine
• Thisislikesolvinganincreasinglydifficultsystemoflinearequations!
A(s) = (1� s)I + sA , 0 s 1
H(s) = B†(s)B(s) , B(s) = P?b A(s)
s = 0
s = 1
|bi
|xi
H(0) = P?b
H(1) = A.P?b .A
![Page 85: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/85.jpg)
Secondalgorithm:An“adiabatic”approachfortheQSLP
• WeassumeforthemomentthatA>1/ 𝜅
• Wedefinetheinterpolatingmatrix
• Similarly,wedefine
• Thisislikesolvinganincreasinglydifficultsystemoflinearequations!
• Theminimumeigenvaluegapisorder1/𝜅2 andthelengthofthepathL islog(𝜅)
A(s) = (1� s)I + sA , 0 s 1
H(s) = B†(s)B(s) , B(s) = P?b A(s)
s = 0
s = 1
|bi
|xi
H(0) = P?b
H(1) = A.P?b .A
L
![Page 86: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/86.jpg)
Therandomizationmethodtoprepareeigenstates[]
s = 0|bi
|xi
H(0) = P?b
H(1) = A.P?b .A
• Byperformingasequenceofprojectivemeasurementsatsufficientlyclosepoints,wecanpreparetherelatedeigenstateswithhighprobability
s1
s2
sq
![Page 87: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/87.jpg)
Therandomizationmethodtoprepareeigenstates[]
s = 0H(0) = P?
b
H(1) = A.P?b .A
• Byperformingasequenceofprojectivemeasurementsatsufficientlyclosepoints,wecanpreparetherelatedeigenstateswithhighprobability
• EachmeasurementcanbesimulatedbyevolvingwiththecorrespondingHamiltonianforrandomtime.Thisreducescoherencesbetweeneigenstates
s1
s2
sq
|bi
|xi
![Page 88: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/88.jpg)
Therandomizationmethodtoprepareeigenstates[]
s = 0H(0) = P?
b
H(1) = A.P?b .A
• Byperformingasequenceofprojectivemeasurementsatsufficientlyclosepoints,wecanpreparetherelatedeigenstateswithhighprobability
• EachmeasurementcanbesimulatedbyevolvingwiththecorrespondingHamiltonianforrandomtime.Thisreducescoherencesbetweeneigenstates
• TheexpectedevolutiontimewiththeHamiltoniansintherandomizationmethodsatisfies
s1
s2
sq
|bi
|xi
TRM = O
✓L2
✏�
◆
![Page 89: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/89.jpg)
Therandomizationmethodtoprepareeigenstates[]
s = 0H(0) = P?
b
H(1) = A.P?b .A
• Byperformingasequenceofprojectivemeasurementsatsufficientlyclosepoints,wecanpreparetherelatedeigenstateswithhighprobability
• EachmeasurementcanbesimulatedbyevolvingwiththecorrespondingHamiltonianforrandomtime.Thisreducescoherencesbetweeneigenstates
• TheexpectedevolutiontimewiththeHamiltoniansintherandomizationmethodsatisfies
s1
s2
sq
|bi
|xi
L isthepathlength𝛥 isthemingap𝜖 istheerror(tracenorm)
TRM = O
✓L2
✏�
◆
![Page 90: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/90.jpg)
TherandomizationmethodfortheQLSP[]
s = 0H(0) = P?
b
H(1) = A.P?b .A
• Byperformingasequenceofprojectivemeasurementsatsufficientlyclosepoints,wecanpreparetherelatedeigenstateswithhighprobability
• EachmeasurementcanbesimulatedbyevolvingwiththecorrespondingHamiltonianforrandomtime.Thisreducescoherencesbetweeneigenstates
• TheexpectedevolutiontimewiththeHamiltoniansintherandomizationmethodsatisfies
s1
s2
sq
|bi
|xi
TRM = O
✓2
log
2()
✏
◆
![Page 91: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/91.jpg)
TherandomizationmethodfortheQLSP[]
• ThestrongdependenceoftheevolutiontimewiththespectralgapsuggestsonetoconsiderotherHamiltoniansthathavethesameeigenstatebutalargereigenvaluegap
![Page 92: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/92.jpg)
TherandomizationmethodfortheQLSP[]
• ThestrongdependenceoftheevolutiontimewiththespectralgapsuggestsonetoconsiderotherHamiltoniansthathavethesameeigenstatebutalargereigenvaluegap
• Forthisproblem,spectralgapamplification[10]isuseful:
H(s) ! H 0(s) = B†(s)⌦ �� +B(s)⌦ �+ =
✓0 B(s)
B†(s) 0
◆
[10]R.D.SommaandS.Boixo,SIAMJ.Comp.593(2013)
![Page 93: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/93.jpg)
TherandomizationmethodfortheQLSP[]
• ThestrongdependenceoftheevolutiontimewiththespectralgapsuggestsonetoconsiderotherHamiltoniansthathavethesameeigenstatebutalargereigenvaluegap
• Forthisproblem,spectralgapamplification[10]isuseful:
• Someresults:
H(s) ! H 0(s) = B†(s)⌦ �� +B(s)⌦ �+ =
✓0 B(s)
B†(s) 0
◆
Let |x(s)i be the eigenstate of 0-eigenvalue of H(s).
Then, |x(s)i|1i is an eigenstate of 0-eigenvalue of H 0(s).
This eigenstate is separated from others by an eigenvalue gap
p�(s)
[10]R.D.SommaandS.Boixo,SIAMJ.Comp.593(2013)
![Page 94: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/94.jpg)
TherandomizationmethodfortheQLSP[]
• ThestrongdependenceoftheevolutiontimewiththespectralgapsuggestsonetoconsiderotherHamiltoniansthathavethesameeigenstatebutalargereigenvaluegap
• Forthisproblem,spectralgapamplificationisuseful:
• Someresults:
• Notethatthepathlengthdidnotchange.TheonlychangefortheRMistheuseofadifferentHamiltonian.
H(s) ! H 0(s) = B†(s)⌦ �� +B(s)⌦ �+ =
✓0 B(s)
B†(s) 0
◆
Let |x(s)i be the eigenstate of 0-eigenvalue of H(s).
Then, |x(s)i|1i is an eigenstate of 0-eigenvalue of H 0(s).
This eigenstate is separated from others by an eigenvalue gap
p�(s)
![Page 95: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/95.jpg)
TherandomizationmethodfortheQLSP[]
• UsingtherandomizationmethodwiththenewHamiltonian,theexpectedevolutiontimeis
TRM = O
✓ log2()
✏
◆
![Page 96: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/96.jpg)
TherandomizationmethodfortheQLSP[]
• UsingtherandomizationmethodwiththenewHamiltonian,theexpectedevolutiontimeis
• Thecaseofnon-positivematrixA canbeanalyzedsimilarlyusing
TRM = O
✓ log2()
✏
◆
A(s) = (1� s)(�anc
z
⌦ I) + s(�anc
x
⌦A)
![Page 97: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/97.jpg)
Therandomizationmethod,theQLSP,andthegatemodel
For A > 0, the Hamiltonian is H 0(s) = (I � |bihb|)((1� s)I + sA)�� +H.c.
![Page 98: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/98.jpg)
Therandomizationmethod,theQLSP,andthegatemodel
For A > 0, the Hamiltonian is H 0(s) = (I � |bihb|)((1� s)I + sA)�� +H.c.
If |bi is sparse and A is sparse, then H 0(s) is also sparse
![Page 99: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/99.jpg)
Therandomizationmethod,theQLSP,andthegatemodel
For A > 0, the Hamiltonian is H 0(s) = (I � |bihb|)((1� s)I + sA)�� +H.c.
If |bi is sparse and A is sparse, then H 0(s) is also sparse
WecanuseaHamiltoniansimulationmethodtobuildaquantumcircuitthatsimulatestheevolution.Thequantumcircuitwillusequeries.
• Thecomplexityintermsofqueriesforis
• ThecomplexityintermsofqueriesforA isalmostorder
|bihb| O(Tb/✏)
CA(/✏, ✏)
![Page 100: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/100.jpg)
Therandomizationmethod,theQLSP,andthegatemodel
For A > 0, the Hamiltonian is H 0(s) = (I � |bihb|)((1� s)I + sA)�� +H.c.
If |bi is sparse and A is sparse, then H 0(s) is also sparse
• Thescalingintheprecisionparametercanbedonepolyligarithmic byusingfastermethodsforeigenpath traversal[11]
WecanuseaHamiltoniansimulationmethodtobuildaquantumcircuitthatsimulatestheevolution.Thequantumcircuitwillusequeries.
• Thecomplexityintermsofqueriesforis
• ThecomplexityintermsofqueriesforA isalmostorder
|bihb| O(Tb/✏)
CA(/✏, ✏)
[11]S.Boixo,E.Knill,andR.D.Somma,arXiv:1005.3034(2010)
![Page 101: Quantum Algorithms for Systems of Linear Equationsqml2018.umiacs.io/slides/Somma.pdf · Let’s consider the problem of solving a system of linear equations or the related problem](https://reader034.vdocuments.us/reader034/viewer/2022050119/5f4f84aaa69fb41f60593269/html5/thumbnails/101.jpg)
Someconclusionsandobservations
• Quantumcomputingseemspromising.Severalquantumalgorithmsforproblemsinlinearalgebrawithsignificantspeedupsexist
• Ipresentedquantumalgorithmstosolvethequantumlinearsystemsproblem.Thetechniquescanbegeneralizedtoapplyotheroperators(otherthantheinverseofamatrix)toquantumstates.
• Theadvantagesofthefirstalgorithmareinthatthecomplexitydependenceonprecisionisonlypolylogarithmic,exponentiallyimprovingpreviousalgorithmsforthisproblem
• Theadvantagesofthesecondalgorithmareinthatitdoesn’trequiremanyancillaryqubitsandtheproblemreducestoasimpleHamiltoniansimulationproblem
• ItwouldbeimportanttounderstandtheapplicabilityofthisalgorithmtoscientificproblemsbeyondtheonesImentioned.Howimportantaretheproblemsandalgorithms?