Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
ProprietaryandConfidential,D-WaveSystemsInc.
Software,CloudServicesandApplications
Copyright©D-WaveSystemsInc. 2
SoftwareDirections
• AlanBaratz newSVPSoftware&Apps
• Newsoftwaretoolsarchitecture– OCEAN
• Moreaccesstosystematmultiplelevels• Microkernel
• Intermediaterepresentations– QUBO
• SubjectMatterExperttools
• ImprovedCloudAccess• 3phaseimplementationstartinginJune
• “Quadrant”MachineLearningSoftware
• Conventionalimplementationtostart
ProprietaryandConfidential,D-WaveSystemsInc. 3
MajorInitiatives
• OceanToolsSuite– PhaseddeliverybeginningDecember,2017– Centraltocloudservice
• CloudServiceBusiness– Goal:Fasteraccessandadeveloperecosystem– Currentlyin4thof8sprints;Alphatestingunderway
• QuadrantMLApplicationsBusiness– Goal:LeverageSOTAgenerativeMLalgorithmstoengagethecommunity– FocusonCancerindicators/treatments,5GSDNs,imagedefectrecognition
• CreativeDestructionLabs(CDL)Participation– FocusedonattractingquantumorientedstartupstoD-Waveplatforms– FourtoptiercompaniesleveragingD-Wave
ProprietaryandConfidential,D-WaveSystemsInc. 4
Ocean1.0ToolsSuite
• GraphMappingandConstraintToolsChains• BeingIntegratedIntoCloudService• InvestigatingAdditionalProblemDecompositionApproaches
ProprietaryandConfidential,D-WaveSystemsInc. 5
ExampleUsage(Python)
importnetworkx asnximportdwave_networkx asdnximportdwave_micro_client_dimod asmicroimportdwave_qbsolv
urlc ='https://cloud.dwavesys.com/sapi'tokenc ='SE-bb7f104b4a99cf9a10eeb9637f0806761c9fcedc'solver_namec ='DW_2000Q_1'
structured_samplerc =micro.DWaveSampler(solver_namec,urlc,tokenc)samplerc =micro.EmbeddingComposite(structured_samplerc)samplerq =dwave_qbsolv.QBSolv()
cloudsi =dnx.structural_imbalance(G,samplerc,num_reads=10000)qbsolvsi =dnx.structural_imbalance(G,samplerq,solver=samplerc)
h={v:node_values[v]forvinG.nodes}J={(u,v):eval foru,vinG.edges}response=samplerc.sample_ising(h,J,num_reads=10000)
ProprietaryandConfidential,D-WaveSystemsInc. 6
CloudServiceBusiness
ServiceComponents
DemosOnlineTraining
CommunitySupport
PaidSupport
KnowledgeBase ProblemStatus
AccountManagement
Open-SourceTools
• NewUI/UXDesign&DevelopmentUnderway• Jupyter HubSelectedforO/LTrainingandProcessing
• Initialnotebookscomplete• SalesforceSelectedforCRM;ZenDesk forSupportandCommunity
• Configurationandintegrationunderway• NewQPUJobSchedulingModelBeingDeveloped
ProprietaryandConfidential,D-WaveSystemsInc. 7
QuadrantMLApplicationsBusiness
• Goal:LeverageSOTAGenerativeMLAlgorithmsforCommunityEngagementandMLApplicationOfferings
• EarlySuccessWorkingwithCustomerstoDevelopApplications– Siemens:UsedCRF-NNtoaccuratelyidentifymedicalinstrumentsinvideos– Huawei:UsedDiscreteDensityEstimationtodetectsleepingcelltowers– ProposaltoNIHforVAEtoidentifygenemarkersforbraindiseases
• LaunchofQuadrant.ai– Pressreleases,website,marketingcampaign
ProprietaryandConfidential,D-WaveSystemsInc. 8
KeyCDLCompanies
• OTILumionics– CreationofnewOLEDmaterials– LeveragingD-Waveforelectronicconfigurationcalculations
• SolidStateAI– ImprovingsemiconductorFAByieldandequipmentfailureprediction– LeveragingD-Waveforstrongclassifierdetermination
• ProteinQure– Newdrugdiscovery– LeveragingD-Wavefor3Dproteinfolding
• AdaptiveFinance– Highyieldequitity trading– LeveragingD-Waveforstrongclassifierdetermination
Copyright©D-WaveSystemsInc. 9
Mission
Tohelpsolvethemostchallengingproblemsinthemultiverse:
• Optimization
• MachineLearning
• MonteCarlo/Sampling
• MaterialScience
Copyright©D-WaveSystemsInc. 10
CustomerApplicationAreas
• Lockheed/USCISI
– SoftwareVerificationandValidation
– Optimization– Aeronautics
– PerformanceCharacterization&Physics
• Google/NASAAmes/USRA
– MachineLearning
– Optimization
– PerformanceCharacterization&Physics
– Research
• LosAlamosNationalLaboratory
– Optimization
– MachineLearning,Sampling
– SoftwareStack
– SimulatingQuantumSystems
– Other(good)Ideas
• CS- 1
– Cybersecurity
• OakRidgeNationalLaboratory
– SimilartoLosAlamos
– MaterialScience&Chemistry
Los Alamos National Laboratory
D-Wave “Rapid Response” Projects (Stephan Eidenbenz, ISTI)
Round 1 (June 2016)1. Accelerating Deep Learning with
Quantum Annealing
2. Constrained Shortest Path Estimation3. D-Wave Quantum Computer as an
Efficient Classical Sampler
4. Efficient Combinatorial Optimization using Quantum Computing
5. Functional Topological Particle Padding6. gms2q—Translation of B-QCQP to
D-Wave7. Graph Partitioning using the D-Wave for
Electronic Structure Problems8. Ising Simulations on the D-Wave QPU9. Inferring Sparse Representations for
Object Classification using the Quantum D-Wave 2X machine
10. Quantum Uncertainty Quantification for Physical Models using ToQ.jl
11. Phylogenetics calculations
Round 2 (December 2016)1. Preprocessing Methods for Scalable Quantum Annealing2. QA Approaches to Graph Partitioning for Electronic
Structure Problems3. Combinatorial Blind Source Separation Using “Ising”4. Rigorous Comparison of “Ising” to Established B-QP
Solution Methods
Round 3 (January 2017)1. The Cost of Embedding2. Beyond Pairwise Ising Models in D-Wave: Searching for
Hidden Multi-Body Interactions3. Leveraging “Ising” for Random Number Generation4. Quantum Interaction of Few Particle Systems Mediated
by Photons5. Simulations of Non-local-Spin Interaction in Atomic
Magnetometers on “Ising”6. Connecting “Ising” to Bayesian Inference Image Analysis7. Characterizing Structural Uncertainty in Models of
Complex Systems8. Using “Ising” to Explore the Formation of Global Terrorist
Networks
Los Alamos National Laboratory 6/27/2017
2016 2017 %CombinatorialOptimization 5 5 10 45%MachineLearning,Sampling 2 2 4 18%UnderstandingDevicePhysics 2 1 3 14%SoftwareStack/Embeddings 1 1 2 9%SimulatingQuantumSystems 2 2 9%Other(good)Ideas 1 1 5%Total 11 11 22 100%
UseCaseTotal
The LANL Rapid Response Project results for 2016 and 2017 are available as PDF’s at:http://www.lanl.gov/projects/national-security-education-center/information-science-technology/dwave/index.php
UNCLASSIFIED Nov. 13, 2017 | 13
ISTI Rapid Response DWave Project (Dan O’Malley):Nonnegative/Binary Matrix Factorization
Imag
e cr
edit:
Lee
&
Seun
g, N
atur
e (1
999)§ Low-rank matrix factorization
• 𝐴 ≈ 𝐵𝐶 where 𝐵%,' ≥ 0 and 𝐶%,' ∈ {0,1}
•
𝐴
≈
𝐵
𝐶
§ Unsupervised machine-learning application• Learn to represent a face as a linear combination of basis images
• Goal is for basis images to correspond to intuitive notions of parts of faces
§ “Alternating least squares”1. Randomly generate a binary 𝐶
2. Solve 𝐵 = argmin6 ∥ 𝐴 − 𝑋𝐶 ∥: classically
3. Solve 𝐶 = argmin6 ∥ 𝐴 − 𝐵𝑋 ∥: on the D-Wave
4. Repeat from step 2
§ Results• The D-Wave NMF approach results in a sparser 𝐶 (85% vs. 13%) and denser but more lossy
compression than the classical NMF approach
• The D-Wave outperforms two state-of-the-art classical codes in a cumulative time-to-target benchmark when a low-to-moderate number of samples are used
UNCLASSIFIED Nov. 13, 2017 | 14
ISTI Rapid Response DWave Project (Hristo Djidev):Efficient Combinatorial Optimization
§ Objectives• Develop D-Wave (DW) algorithms for NP-hard
problemsFocus: the max clique (MC) problem:
• Study scalability/accuracy issues and ways to mitigate them
• Characterize problem instances for which D-Wave may outperform classical alternatives
§ Results• The MC problem can be solved accurately
and fast on DWo but so can classical methodso no quantum advantage for typical
problem instances fitting DW (of size ~45)
• Running on larger (Chimera-like) graphso Chimera graph is
modified by merging a set of randomlyselected edges intoa vertex
o Resulting graphs ofsizes upto 1000 are used as inputs to MC problem
o DW beats simulated annealing, its classical analogue, by a factor of more than 106
o In order to see a quantum advantage for the MC problem, graph sizes should be > 300
o In order to see a quantum advantage for the MC problem, graph sizes should be > 300
orig.graph cliqueofmaxsize prob edges: 3068, Energy prob upper bound: 3930.5
-2 2 Vertices
-1 1 Couplers
Solution : -1 +1
200 400 600 800 1000
−50
−40
−30
−20
−10
0
Graph size
best
cliq
ue s
ize fo
und,
rela
tive
to D
wave
2X
Dwave 2XPPHaSA−clique fastSA−clique slowSA−isingfmc
0 200 400 600 800 1000
1e−0
21e
+00
1e+0
21e
+04
1e+0
6
Graph size
spee
dup
Quality comparison Speed comparison
UNCLASSIFIED Nov. 13, 2017 | 15
ISTI Rapid Response DWave Project (Sue Mniszewski): Quantum Annealing Approaches to Graph Partitioning for Electronic Structure Problems
§ Motivated by graph-based methods for quantum molecular dynamics (QMD) simulations
§ Explored graph partitioning/clustering methods formulated as QUBOs on D-Wave 2X
§ Used sapi and hybrid classical-quantum qbsolv software tools
§ Comparison with state-of-the-art tools§ High-quality results on benchmark (Walshaw),
random, and electronic structure graphsGraph N Best METIS KaHIP qbsolv
Add20 2395 596 723 613 602
3elt 4720 90 91 90 90
Bcsstk33 8738 10171 10244 10171 10171Minimize edge counts between 2 parts on Walshaw graphs.
k-Concurrent Partitioning for Phenyl Dendrimer.
k-parts METIS qbsolv2 705 706
4 20876 2648
8 22371 15922
16 28666 26003
k-Concurrent clustering for IGPS Protein Structure:resulting 4 communities share common sub-structure. Comparable to classical methods.
UNCLASSIFIED Nov. 13, 2017 | 16
ISTI Rapid Response Project (Carleton Coffrin):Challenges and Successes of Solving Binary Quadratic Programming Benchmarks on the DW2X QPU
§ Looking to the Future• I have drunk D-Wave Kool-Aid• RAN1 convinced me that the DW2X has huge potential• I believe, in the next 5 years, QPU’s will be very disruptive to
optimization research
Biswas, SMC-IT, 28 Sept 2017
Quantum Computing for NASA Applications
17
Data Analysisand Data Fusion
AirTrafficManagement
Mission Planning, Scheduling, and Coordination
V&V and Optimal Sensor
Placement
Topologically-aware Parallel
Computing
Anomaly Detection and
Decision Making
Common Feature: Intractable problems on classical supercomputers
Objective: Find “better” solution• Faster• More precise• Not found by classical algorithm
Current NASA Research in Annealing Applications
Machine Learning
Graph-based Fault Detection
Complex Planning and Scheduling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1617
18
19
20
21
22
23
24
25
26
27
28
29
30
31
3233
34
35
36
37
38
39 4041 42
43
44
45
46
• General Planning Problems (e.g., navigation, scheduling, asset allocation) can be solved on a quantum annealer (such as D-Wave)
• Developed a quantum solver for Job Shop Scheduling that pre-characterizes instance ensembles to design optimal embedding and run strategy – tested at small scale (6x6) but potentially could solve intractable problems (15x15) with 10x more qubits
• Analyzed simple graphs of Electrical Power Networks to find the most probable cause of multiple faults – easy and scalable QUBO mapping, but good parameter setting (e.g., gauge selection) key to finding optimal solution –now exploring digital circuit Fault Diagnostics
• Boltzmann sampling commonly used in Machine Learning, particularly Deep Learning. Quantum computing has provable advantage for some sampling problems. Demonstrated learning when using a QA as a Boltzmann sampler.
01 0 11 1 10 01 0 0 01 0 0
CircuitBreakers
Sensors
Observations
IN: configs.
OUT: params.
QA {J , h}
D-Wave run results: established baseline performance for QA on these applications
Scheduling Applications
Job-Shop scheduling: Complete quantum-classical solver framework with pre-
processing, compilation/run strategies, decomposition methods
D. Venturelli, D. J.J. Marchand, G. Rojo, Quantum Annealing Implementation of Job Shop Scheduling, arXiv:1506.08479
Eleanor G. Rieffel, Davide Venturelli, Minh Do, Itay Hen, Jeremy Frank, ParametrizedFamilies of Hard Planning Problems from Phase Transitions, AAAI-14.E. G. Rieffel, D. Venturelli, B. O'Gorman, M. B. Do, E. Prystay, V.N. Smelyanskiy, A case study in programming a quantum annealer for hard operational planning problems, Q. Information Processing, 14, (2014)
Comparison with state-of-the-art application-specific algorithms:current best planners
Scheduling problems as testbed for resource-bounded tailored embedding methods
10 20 30 40 50
Problem size n: number of tasks
Med
ian
Run
time
[sec
]
0.01
0.1
1
10
100
1000
10,000
Planner Comparison: All Scheduling Problems
FF:α=1.11 ± 0.061LPG: α=0.69 ± 0.139M: α=0.1 ± 0.007Mp: α=0.54 ± 0.035
Solved problems with 6 machines and 6 jobs: analyzed scaling of tractability
Graph coloring
Mars Lander activity scheduling
Airport runwayscheduling
• T. Tran, M. Do, E. Rieffel, J. Frank, Z. Wang, B. O'Gorman, D. Venturelli, J. Beck, A Hybrid Quantum-Classical Approach to Solving Scheduling Problems, SOCS’16
• T. Tran, Z. Wang, M. Do, E. Rieffel, J. Frank, B. O'Gorman, D. Venturelli, J. Beck, Explorations of Quantum-Classical Approaches to Scheduling a Mars Lander Activity Problem, Workshops AAAI’16
QA-guided tree search
A. Perdomo-Ortiz et al., On the readiness of quantum optimization machines for industrial applications arXiv:1708.09780
Fault DiagnosisFirst comprehensive study addressing the readiness of
quantum annealing for real-world applicationsSix different algorithms (SA, PT-ICM, QMC, SAFARI, SAT-based, and DWave2X)In all three problem Hamiltonian representations (PUBO, QUBO, Chimera)
Addressed future quantum annealer design for quantum advantage in applications with practical relevance
• What is the impact of higher-order terms? • Need for non-stoquastic Hamiltonians? • Impact of connectivity? …
Compresseddata
i~d⇢✓(t)dt
= [H✓, ⇢✓]i~d⇢✓(t)
dt= [H✓, ⇢✓]
✓
Rawinputdata
Quantumsampling
Measurement
Hiddenlayers
ClassicalgenerationorreconstructionofdataIn
ference
Qua
ntum
processing
Classicalpre-a
ndpost-p
rocessing
Visibleunits Hiddenunits Qubits
Trainingsamples
Generatedsamples
• Hybridproposalthatworksdirectlyonalow-dimensionalrepresentationofthedata.Newparadigm:UsedeeplearningtoassistQMLimplementationinnear-termQC
Anear-termapproachforquantum-enhancedmachinelearning
Benedetti,Realpe-Gomez,andPerdomo-Ortiz.Quantum-assistedHelmholtzmachines:Aquantum-classicaldeeplearningframeworkforindustrialdatasetsinnear-termdevices.arXiv:1708.09784 (2017).
Feasibility study: Using quantum-classical hybrids to assure the availability of the UAS Traffic Management (UTM) network
against communication disruptions
Kopardekar, P., Rios, J., et. al., Unmanned Aircraft System Traffic Management (UTM) Concept of Operations, DASC 2016
Future • Higher vehicle density• Heterogeneous air vehicles• Mixed equipage• Greater autonomy• More vulnerability to
communications disruptions
Newly funded effort in aeronautics
Explore quantum approaches to• Robust network design• Track and locate of a moving jammer • Secure communication of codes
supporting anti-jamming protocols
30 month effort: harness the power of quantum computing and communication to address the cybersecurity challenge of availability
Joint with NASA Glenn, who are working on QKD for spread spectrum codes
Prior work (NASA-DLR collaboration): T. Stollenwerk et al., Quantum Annealing Applied to De-Conflicting Optimal Trajectories for Air Traffic Management
Copyright©D-WaveSystemsInc. 23
Mission
Tohelpsolvethemostchallengingproblemsinthemultiverse:
• Optimization
• MachineLearning
• MonteCarlo/Sampling
• MaterialScience
Quantum Computing at Volkswagen:Traffic Flow Optimization using the D-Wave Quantum Annealer
D-Wave Users Group Meeting - National Harbour, MD 27.09.2017 – Dr. Gabriele Compostella
The Question that drove us …
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 25
Is there a real-world problem that could be addressed with a
Quantum Computer?
YES: Traffic flow optimisation
Everybody knows traffic (jam) and normally nobody likes it.Image courtesy of think4photop at FreeDigitalPhotos.net
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 26
Public data set: T-Drive trajectory
https://www.microsoft.com/en-us/research/publication/t-drive-trajectory-data-sample/
Beijing• ~ 10.000 Taxis• 2.2. – 8.2.2008
data example:
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 27
Result: unoptimised vs optimised traffic
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 28
Volkswagen Quantum Computing in the news
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 29
27.09.2017 K-SI/LD | Dr. Gabriele Compostella 30
HETEROGENEOUS QUANTUM COMPUTING FOR SATELLITE OPTIMIZATIONGID EON BAS S
BOOZ AL L EN HAM ILTON
September 2017
BO O Z AL L EN • DI G I T A L
Booz Allen Hamilton Restricted, Client Proprietary, and Business Confidential.
Traveling Salesman
Vehicle Routing
Logistics
Circuit Design
Network Design
Manufacturing
Machine Learning
Artificial Intelligence
Robotics
System DesignOptimization
Combinatorial Chemistry
Drug Discovery
QUANTUMANNEALING HAS MANY REAL-WORLDAPPLICATIONS
BO O Z AL L EN • DI G I T A L
Booz Allen Hamilton Restricted, Client Proprietary, and Business Confidential.
CONCLUSIONS
18
+ As problems and datasets grow, modern computing systems have had to scale with them. Quantum computing offers a totally new and potentially disruptive computing paradigm.
+ For problems like this satellite optimization problem, heterogeneous quantum techniques will be required to solve the problem at larger scales.
+ Preliminary results on this problem using heterogeneous classical/quantum solutions are very promising.
+ Exploratory studies in this area have the potential tobreak new ground as one of the first applications ofquantum computing to a real-world problem
Display Advertising Optimization by
Quantum Annealing Processor
Shinichi Takayanagi*, Kotaro Tanahashi*, Shu Tanaka†*Recruit Communications Co., Ltd.
† Waseda University, JST PRESTO
(C)Recruit Communications Co., Ltd.
Behind the Scenes
35
DSPSSP
RTB
AdvertiserPublisher
Impression
SSP: Supply-Side PlatformDSP: Demand-Side PlatformRTB: Real Time Bidding
AD
AD
AD
1.0$
0.9$
0.7$
Winner!
(C)Recruit Communications Co., Ltd.
CTR Prediction with Machine-Learning
• Machine-Learning (ML) tech is often used for CTR prediction
• ML has succeeded in this field
36
Click F1 F2 F3
1 M 01 2.13
0 F 07 2.12
0 F 23 4.2
? F 99 1.2
Click or NotPrediction
Matrix expressionUsers
Model
(Click-through-rate)
(C)Recruit Communications Co., Ltd.
0
4
8
12
16
0 6 12 18
Budget Pacing
• Budget pacing is also important
• Control of budget pacing helps
advertisers to…
– Reach a wider range of audience
– Avoid a premature campaign stop / overspending
37
Too fast
Budget pacing controlled
Budg
et
spendi
ng
Target budget
Time (hours)
(C)Recruit Communications Co., Ltd.
4. Summary
• Budget pacing is important for display advertising• Formulate the problem as QUBO• Use D-Wave 2X to solve budget pacing control
optimization problem• Quantum annealing finds a better solution than the
greedy method.
38
Copyright©D-WaveSystemsInc. 39
DENSOOptimizationProjects
VideosfromCES,LasVegas,January2018
AutonomousDrivinghttps://www.youtube.com/watch?v=Bx9GLH_GklA
FactoryOptimizationhttps://www.youtube.com/watch?v=BkowVxTn6EU
Copyright©D-WaveSystemsInc. 40
Mission
Tohelpsolvethemostchallengingproblemsinthemultiverse:
• Optimization
• MachineLearning
• MonteCarlo/Sampling
• MaterialScience
41 billings7893
ORNL is managed by UT-Battelle for the US Department of Energy
42 Presentation_name
Adiabatic Quantum Programming at ORNL: Workflow Environments and HPC Integration APIs
The First:
There are currently 3 main challenges in Deep Learning
A Study of Complex Deep Learning Networks on High Performance, Neuromorphic, and Quantum Computers
QuantumMachineLearningforElection Modelling
Election2016:Casestudyinthedifficultlyof sampling
Wheredidthemodelsgo wrong?
QuantumMachineLearningforElectionModelling– MaxHenderson, 2017 44
Forecastingelectionsonaquantum computer
QuantumMachineLearningforElectionModelling– MaxHenderson, 2017 45
• Quantumcomputingresearchhasshownpotentialbenefits(speedups)intrainingvariousdeepneural networks1-3
• Coreidea:UseQC-trainedmodelstosimulateelectionresults.Potential benefits:
• Moreefficientsampling/ training• Intrinsic,tuneablestate correlations• Inclusionofadditionalerror models
1. Adachi, Steven H., and Maxwell P. Henderson. "Application of quantum annealing totraining of deep neural networks." arXiv preprint arXiv:1510.06356 (2015).2. Benedetti,Marcello,etal."Estimationofeffectivetemperaturesinquantumannealersforsamplingapplications:Acasestudywithpossibleapplicationsindeep
learning."PhysicalReviewA94.2(2016): 022308.3. Benedetti, Marcello, et al. "Quantum-assisted learning of graphical models with arbitrary pairwise connectivity." arXiv preprint arXiv:1609.02542 (2016).
Summary
QuantumMachineLearningforElectionModelling– MaxHenderson, 2017 46
• TheQC-trainednetworkswereabletolearnstructureinpollingdatatomakeelectionforecastsinlinewiththemodelsof 538
• Additionally,theQC-trainednetworksgaveTrumpamuchhigherlikelihoodofvictoryoverall,eventhoughthestate’sfirstordermomentsremained unchanged
• Ideallyinthefuture,wecouldrerunthismethodusingcorrelationsknownwithmoredetailin-housefor 538
• Finally,theQC-trainednetworkstrainedquickly,andsinceeachmeasurementisasimulation,eachiterationofthetrainingmodelproduced25,000simulations(oneforeachnationalerrormodel),whichalreadyeclipsesthe20,000simulations538performseachtimetheyreruntheir models
COPYRIGHT 2016 LOCKHEED MARTIN CORPORATION – ALL RIGHTS RESERVED47
Quantum Enabled Machine Learning
Supervised Learning: Improving Neural Network Training
Adachi, Steven H., and Maxwell P. Henderson. "Application of Quantum Annealing to Training of Deep Neural Networks." arXiv preprint arXiv:1510.06356 (2015).
visiblelayer
hiddenlayer vi
sibl
e no
de 1
hidden node 1
hidden node 2
hidden node 3
hidden node 4
hidden node 5
hidden node 6
hidden node 7
hidden node 8
visi
ble
node
2
visi
ble
node
3
visi
ble
node
4
visi
ble
node
5
visi
ble
node
6
visi
ble
node
7
visi
ble
node
8
AlejandroPerdomo-OrtizSeniorResearchScientist,QuantumAILab.atNASAAmesResearchCenterandatthe
UniversitySpaceResearchAssociation,USAHonorarySeniorResearchAssociate,ComputerScienceDept.,UCL,UK
NationalHarbor,MD,September28,2017
Opportunitiesandchallengesinquantum-enhancedmachinelearninginnear-termquantumcomputers
QUBITSD-waveUserGroup2017
Funding:
Perdomo-Ortiz,Benedetti,Realpe-Gomez,andBiswas.arXiv:1708.09757 (2017).ToappearintheQuantumScienceandTechnology(QST)invitedspecialissueon“Whatwouldyoudowitha1000qubit device?”
Los Alamos National Laboratory
28-Sep-2016 | 49
Copyright©D-WaveSystemsInc. 50
Mission
Tohelpsolvethemostchallengingproblemsinthemultiverse:
• Optimization
• MachineLearning
• MonteCarlo/Sampling
• MaterialScience
Copyright©D-WaveSystemsInc. 51
Z(2)latticegaugetheoryKosterlitz-Thouless model
3Dtransverse-fieldIsing model
QuantumMaterialScience@D-Wave
R.Harris
A.King E.Dahl
TheFuture,Maybe
• 2018Predictions• Beyond
Copyright©D-WaveSystemsInc. 53
GateModelMachines- 2018
• ~50– 100qubitmodelsrunning• Nolargescaleerrorcorrection• NoisyIntermediateScaleQC’s(NISQ)*• Knowifsomeproblemswillrunwithout
errorcorrection• QuantumMaterialScience?• NoShor’sAlgorithm• Quantum“Supremacy”perhapsfor
syntheticbenchmark• Importanceoferrorcorrectionandpotentialappsbecomesclear*“QuantumComputingintheNISQeraandbeyond”,JohnPreskill,CalTech,arXiv:1801.00862
Copyright©D-WaveSystemsInc. 54
.
• 75– 100“proto-apps”on2000QD-WaveSystem
• ~Halfapproachingclassicalperformanceon
smallishproblems
• DemonstrateQuantumMaterialScience
breakthrough
• Quantum“Advantage”demonstrations
• IARPAQEOandD-Wavehighercoherence
qubitdemonstrations
• Trajectoryto4000-5000qubitsystem,better
connectivity,lowernoise
15mK
QuantumAnnealing- 2018
Copyright©D-WaveSystemsInc. 55
AndBeyond
• BiginvestmentsinQC– China $11B
– EUFlagship $1B+
– UKHubs1&2 <$1B
– Japan ChristmasDaymeeting
• U.S.andCanada– fragmented– 2019U.S.budgetproposal– DOE$100M,NSF$30M,others?
• QuantumDiversity• Moresmartpeopleworkingonappsandsoftwaretools
• Bo’sUnifiedTheoryofQuantumComputing
Copyright©D-WaveSystemsInc. 56
President’sNationalStrategicComputingInitiative
Copyright©D-WaveSystemsInc. 57
AfterNike™
QuantumComputingneedsyouto:
JustDoIt™Probably
Copyright©D-WaveSystemsInc. 58
ForMoreInformationSee
D-WaveUsersGroupPresentations:– https://dwavefederal.com/qubits-2016/– https://dwavefederal.com/qubits-2017/
LANLRapidResponseProjects:– http://www.lanl.gov/projects//national-security-education-center/information-science-technology/dwave/index.php