Machine learning with quantum computers

Speck&Tech
Speck&TechSpeck&Tech
Machine Learning with Quantum Computers
Speck&Tech event
Davide Pastorello
Department of Mathematics, University of Bologna
and
Trento Institute for Fundamental Physics and Applications, INFN
Impact Hub Trento. September 21, 2023
Picture credit: IBM
Premise: The World is quantum
At microscopic length scales (molecules, atoms, electrons,...): Newton’s laws of
dynamics and electromagnetism based on Maxwell’s equations do not work.
We need quantum mechanics to describe physical phenomena.
Quantum things:
• Superposition of states
(is the electron either "here" or "there"? In both places!).
• Observations alters the state of the observed system.
(We observe the electron "here" with some probability)
• Entanglement ("spooky action at distance" as Einstein said)
Using physical phenomena to process information is a good and old idea...
What if are we able to use quantum phenomena?
A bit of history...
1981. At the First Conference of Physics of Computation (MIT), R. Feynman observed
that quantum systems cannot be efficiently simulated by classical computers and
proposed a model of quantum computer.
1985. D. Deutsch described the universal quantum computer in terms of the quantum
Turing machine.
1994. P. Shor proposed an algorithm for a quantum computer to factor large integers
in polynomial time.
1995. L. Grover presents the quantum search algorithm in an unsorted database.
2000s. First prototypes of working quantum computers.
2010s. Quantum computers on the market (e.g. D-Wave, IBM Q System 1).
Present days (2023)
• Anyone can run his/her quantum algorithms on available quantum machines (a
basic access is usually free).
• Real-life problems are tackled by quantum computers.
• The interest by industry and non academic institutions is dramatically increasing.
• Availability of fault-tolerant, error-corrected, universal quantum computers?
Not yet, we are in the Noisy Intermediate-Scale Quantum (NISQ).
• Quantum supremacy? Not yet.
• Hot topics: quantum optimization and quantum machine learning...
Quantum computers exist
Examples of quantum computers:
D-Wave IBM Quantum
Quantum bits
Examples of qubits:
• Particles with spin 1/2;
• Polarized Photons;
• Controlled superconducting circuits.
Qubits (mathematical description)
Quantum state superposition
The state of a qubit is a unit vector* in C2:
| i = ↵|0i + |1i ↵, 2 C , |↵|2
+ | |2
= 1
{|0i, |1i} computational basis.
A measurement process affects the qubit state
Measurement of a qubit (e.g. we measure the polarization of a photon)
P(0) = |↵|2
P(1) = | |2
The qubit state after the measurement is |ii if the outcome is i = 0, 1.
*The vectors | i and ei✓
| i with ✓ 2 R represent the same physical state.
Qubits (mathematical description)
Tensor product
| i =
✓
↵
◆
2 C2
|'i =
✓ ◆
2 C2
| i ⌦ |'i =
0
B
B
@
↵
↵
1
C
C
A 2 C4
2 qubits (as a composite system) are described in:
C2
⌦ C2
:= span{| i ⌦ |'i : | i, |'i 2 C2
} = C4
.
n qubits (as a composite system) are described in:
(C2
)⌦n
= C2n
.
A new kind of information
Empirical evidence:
Quantum randomness, state superpositions, entanglement are
physical phenomena not simply theoretical interpretations.
+
Encoding information into qubits (or more general quantum objects) allows completely
new kinds of:
• information storing and processing
• data representation
• telecommunication
• cybersecurity
• ...
In particular we are interested in machine learning!
Quantum gates
Let {|0i, |1i} be the computational basis
• Hadamard gate:
|xi H
1
p
2
(|0i + ( 1)x
|1i) x 2 {0, 1}
• Phase gate:
↵|0i + |1i P ↵|0i + ei
|1i
• CNOT gate:
|xi • |xi
|yi |x yi
x, y 2 {0, 1}
Theorem: {H, P⇡/4, CNOT} is a universal set for quantum computation.
Machine learning with quantum computers?
Achievements in Quantum Machine Learning (since 2013)
Algorithm Quantum speedup
K-medians Quadratic
Hierarchical clustering Quadratic
K-means Exponential
Principal component analysis Exponential
Support vector machines Exponential
Several classifiers Exponential
Neural networks ?
General observation
Quantum architectures enable new paradigms of data representation and information
processing.
Current challenge
Devising quantum learning mechanisms for available or near-term quantum machines.
A quantum binary classifier
The classification model
Assume data are represented in a feature space.
The model assigns a binary label to a new data instance evaluating its similarity to any training
point.
Quantum implementation on the IBM ibmq_melbourne
• Quantum superposition of training vectors;
• Test instance in quantum superposition of the two classes;
• Parallel computing over the superposed data.
Classical complexity: O(Nd)
Quantum complexity: O(log(Nd))
Reference:
D. Pastorello, E. Blanzieri. A quantum binary classifier based on cosine similarity
Proc. IEEE International Conference on Quantum Computing and Engineering (QCE 21)
Some QML schemes designed for universal quantum computers
• Quantum divisive clustering
W. Aïmeur et al. Quantum clustering algorithms ICML ’07: Proceedings of the 24th
international conference on Machine learning (2007)
• Quantum principal component analysis
S. Lloyd, et al. Quantum principal component analysis Nature Physics 10, 631 (2014)
• Quantum support vector machine
P. Rebentrost et al. Quantum support vector machine for big data classification Phys. Rev. Lett.
113, 130503 (2014)
• Quantum nearest neighbor
N. Wiebe et al. Quantum Algorithms for Nearest-Neighbor Methods for Supervised and
Unsupervised Learning Quantum Information and Computation 15(3,4): 0318- 0358 (2015)
• Quantum perceptron M. Schuld et al. Simulating a perceptron on a quantum computer
Physics Letters A, 379, pp. 660-663 (2015)
• Quantum meta-learning M. Wilson et al. Optimizing quantum heuristic with meta-learning
Quantum Machine Intelligence 3 (2021)
Quantum Annealing
Quantum Annealers
The hardware is a quantum spin glass, i.e. a collection of qubits arranged in the vertices of a
graph (V , E) where edges represent the interactions between neighbors.
Example: D-Wave Chimera topology
Quantum Annealing
Annealing process (annealing time ⇠ 20µs)
By energy dissipation the quantum system evolves in the ground state (the less energetic
state) corresponding to the solution of a given optimization problem.
Quantum Annealing
Shortcoming
Representing a given problem into the quantum hardware architecture is difficult in
general and can destroy the efficiency of the quantum computation.
Proposed solution
Quantum Annealing Learning Search (QALS):
A hybrid quantum-classical algorithm enabling a learning mechanism so that the
quantum machine learns the representation of the problem on its own without an
expensive classical pre-processing.
References:
-) Davide Pastorello, E. Blanzieri. Quantum Annealing Learning Search for solving QUBO problems.
Quantum Information Processing 18: 303 (2019)
-) A. Bonomi et al. Quantum annealing learning search implementations in Quantum Information and
Computation, v. 22, n. 3&4 (2022)
ML with quantum annealers
• Boltzmann machine implementation
M.H. Amin Quantum Boltzmann machine Phys. Rev. X 8, 021050 (2018)
• Classification
N. T. Nguyen et al. Image classification using quantum inference on the D-wave 2X. In: 2018
IEEE International Conference on Rebooting Computing (ICRC), pp. 1-7. IEEE (2018)
• Clustering
V. Kumar et al. Quantum Annealing for Combinatorial Clustering Quantum Inf Process (2018)
17: 39
• Training of a SVM
D. Willsch et al. Support vector machines on the D-wave quantum annealer Computer Physics
Communications 248, 107006 (2020)
...and quantum neural networks?
A feedforward QNN made by single-qubit rotations and CNOTs:
A QNN of this form achieves higher values of effective dimension w.r.t. to a classical NN
with same number of parameters.
Ref.: A. Abbas et al. The power of quantum neural networks Nat. Comput. Sci. 1, 403-409 (2021)
Some concluding remarks
• Quantum architectures enable new kinds of data processing.
• There is a bottleneck about the efficient data representation into quantum hardware
(QRAM are not feasible yet).
• The era of quantum machine learning is just started but it is maybe the most promising
road for putting forward the frontiers of quantum computing.
• The quantum-classical hybrid approach is the best way to use the existing quantum
machines.
• Providing efficient quantum algorithms is a very difficult task.
So we have to devise learning mechanisms of new quantum algorithms towards a
"quantum meta-learning"...
In other words, quantum computers must learn how to solve problems on their own...
Thanks for your attention!
Picture credit: IBM
1 de 20

Recomendados

Quantum computing por
Quantum computingQuantum computing
Quantum computingKrishna Patel
6.1K visualizações21 slides
Quantum Information FINAL.pptx por
Quantum Information FINAL.pptxQuantum Information FINAL.pptx
Quantum Information FINAL.pptxgitrahekno
17 visualizações36 slides
Quantum Computing Applications in Power Systems por
Quantum Computing Applications in Power SystemsQuantum Computing Applications in Power Systems
Quantum Computing Applications in Power SystemsPower System Operation
298 visualizações9 slides
Macaluso antonio meetup dli 2020-12-15 por
Macaluso antonio  meetup dli 2020-12-15Macaluso antonio  meetup dli 2020-12-15
Macaluso antonio meetup dli 2020-12-15Deep Learning Italia
31 visualizações32 slides
Quantum & AI in Finance por
Quantum & AI in FinanceQuantum & AI in Finance
Quantum & AI in FinanceObject Automation
289 visualizações42 slides
Quantum Computing Lecture 1: Basic Concepts por
Quantum Computing Lecture 1: Basic ConceptsQuantum Computing Lecture 1: Basic Concepts
Quantum Computing Lecture 1: Basic ConceptsMelanie Swan
1K visualizações61 slides

Mais conteúdo relacionado

Similar a Machine learning with quantum computers

Seminar por
SeminarSeminar
SeminarRitikesh Bhaskarwar
570 visualizações25 slides
An Introduction to Quantum computing por
An Introduction to Quantum computingAn Introduction to Quantum computing
An Introduction to Quantum computingJai Sipani
302 visualizações20 slides
Quantum computing in machine learning por
Quantum computing in machine learningQuantum computing in machine learning
Quantum computing in machine learningkhalidhassan105
527 visualizações19 slides
177 por
177177
177vivatechijri
13 visualizações6 slides
Quantum computing seminar por
Quantum computing seminarQuantum computing seminar
Quantum computing seminarPankaj Kumar
15.3K visualizações33 slides
Quantum communication and quantum computing por
Quantum communication and quantum computingQuantum communication and quantum computing
Quantum communication and quantum computingIOSR Journals
621 visualizações5 slides

Similar a Machine learning with quantum computers(20)

An Introduction to Quantum computing por Jai Sipani
An Introduction to Quantum computingAn Introduction to Quantum computing
An Introduction to Quantum computing
Jai Sipani302 visualizações
Quantum computing in machine learning por khalidhassan105
Quantum computing in machine learningQuantum computing in machine learning
Quantum computing in machine learning
khalidhassan105527 visualizações
177 por vivatechijri
177177
177
vivatechijri13 visualizações
Quantum computing seminar por Pankaj Kumar
Quantum computing seminarQuantum computing seminar
Quantum computing seminar
Pankaj Kumar15.3K visualizações
Quantum communication and quantum computing por IOSR Journals
Quantum communication and quantum computingQuantum communication and quantum computing
Quantum communication and quantum computing
IOSR Journals621 visualizações
QuantumComputersPresentation por Vinayak Suresh
QuantumComputersPresentationQuantumComputersPresentation
QuantumComputersPresentation
Vinayak Suresh1K visualizações
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM por csitconf
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
csitconf495 visualizações
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM por cscpconf
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHMCOMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
COMPUTATIONAL PERFORMANCE OF QUANTUM PHASE ESTIMATION ALGORITHM
cscpconf35 visualizações
Quantum Computing por Mehdi Rezaie
Quantum ComputingQuantum Computing
Quantum Computing
Mehdi Rezaie1.6K visualizações
Quantum computers por Nolesh_Warke
Quantum computersQuantum computers
Quantum computers
Nolesh_Warke2.4K visualizações
Quantum computers por Rishabh Jindal
Quantum computersQuantum computers
Quantum computers
Rishabh Jindal2.2K visualizações
Quantum Blockchains por Melanie Swan
Quantum BlockchainsQuantum Blockchains
Quantum Blockchains
Melanie Swan1.1K visualizações
Mathematical_Introduction_to_Quantum_Computation por Brian Flynn
Mathematical_Introduction_to_Quantum_ComputationMathematical_Introduction_to_Quantum_Computation
Mathematical_Introduction_to_Quantum_Computation
Brian Flynn206 visualizações
Algoritmo quântico por XequeMateShannon
Algoritmo quânticoAlgoritmo quântico
Algoritmo quântico
XequeMateShannon257 visualizações
Fundamentals of quantum computing part i rev por PRADOSH K. ROY
Fundamentals of quantum computing   part i revFundamentals of quantum computing   part i rev
Fundamentals of quantum computing part i rev
PRADOSH K. ROY101 visualizações
Quantum Research Summary por Brenda Torres
Quantum Research SummaryQuantum Research Summary
Quantum Research Summary
Brenda Torres3 visualizações
Let's build a quantum computer! por Andreas Dewes
Let's build a quantum computer!Let's build a quantum computer!
Let's build a quantum computer!
Andreas Dewes2.8K visualizações

Mais de Speck&Tech

Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ... por
Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...
Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...Speck&Tech
68 visualizações12 slides
Monitorare una flotta di autobus: architettura di un progetto di acquisizione... por
Monitorare una flotta di autobus: architettura di un progetto di acquisizione...Monitorare una flotta di autobus: architettura di un progetto di acquisizione...
Monitorare una flotta di autobus: architettura di un progetto di acquisizione...Speck&Tech
66 visualizações18 slides
Why LLMs should be handled with care por
Why LLMs should be handled with careWhy LLMs should be handled with care
Why LLMs should be handled with careSpeck&Tech
84 visualizações21 slides
Privacy in the era of quantum computers por
Privacy in the era of quantum computersPrivacy in the era of quantum computers
Privacy in the era of quantum computersSpeck&Tech
203 visualizações37 slides
Give your Web App superpowers by using GPUs por
Give your Web App superpowers by using GPUsGive your Web App superpowers by using GPUs
Give your Web App superpowers by using GPUsSpeck&Tech
207 visualizações60 slides
From leaf to orbit: exploring forests with technology por
From leaf to orbit: exploring forests with technologyFrom leaf to orbit: exploring forests with technology
From leaf to orbit: exploring forests with technologySpeck&Tech
101 visualizações35 slides

Mais de Speck&Tech(20)

Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ... por Speck&Tech
Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...
Home4MeAi: un progetto sociale che utilizza dispositivi IoT per sfruttare le ...
Speck&Tech68 visualizações
Monitorare una flotta di autobus: architettura di un progetto di acquisizione... por Speck&Tech
Monitorare una flotta di autobus: architettura di un progetto di acquisizione...Monitorare una flotta di autobus: architettura di un progetto di acquisizione...
Monitorare una flotta di autobus: architettura di un progetto di acquisizione...
Speck&Tech66 visualizações
Why LLMs should be handled with care por Speck&Tech
Why LLMs should be handled with careWhy LLMs should be handled with care
Why LLMs should be handled with care
Speck&Tech84 visualizações
Privacy in the era of quantum computers por Speck&Tech
Privacy in the era of quantum computersPrivacy in the era of quantum computers
Privacy in the era of quantum computers
Speck&Tech203 visualizações
Give your Web App superpowers by using GPUs por Speck&Tech
Give your Web App superpowers by using GPUsGive your Web App superpowers by using GPUs
Give your Web App superpowers by using GPUs
Speck&Tech207 visualizações
From leaf to orbit: exploring forests with technology por Speck&Tech
From leaf to orbit: exploring forests with technologyFrom leaf to orbit: exploring forests with technology
From leaf to orbit: exploring forests with technology
Speck&Tech101 visualizações
Innovating Wood por Speck&Tech
Innovating WoodInnovating Wood
Innovating Wood
Speck&Tech141 visualizações
Behind the scenes of our everyday Internet: the role of an IXP like MIX por Speck&Tech
Behind the scenes of our everyday Internet: the role of an IXP like MIXBehind the scenes of our everyday Internet: the role of an IXP like MIX
Behind the scenes of our everyday Internet: the role of an IXP like MIX
Speck&Tech99 visualizações
Architecting a 35 PB distributed parallel file system for science por Speck&Tech
Architecting a 35 PB distributed parallel file system for scienceArchitecting a 35 PB distributed parallel file system for science
Architecting a 35 PB distributed parallel file system for science
Speck&Tech127 visualizações
Truck planning: how to certify the right route por Speck&Tech
Truck planning: how to certify the right routeTruck planning: how to certify the right route
Truck planning: how to certify the right route
Speck&Tech228 visualizações
Break it up! 5G, cruise control, autonomous vehicle cooperation, and bending ... por Speck&Tech
Break it up! 5G, cruise control, autonomous vehicle cooperation, and bending ...Break it up! 5G, cruise control, autonomous vehicle cooperation, and bending ...
Break it up! 5G, cruise control, autonomous vehicle cooperation, and bending ...
Speck&Tech229 visualizações
AI and Space: finally, no more arguing with the GPS por Speck&Tech
AI and Space: finally, no more arguing with the GPSAI and Space: finally, no more arguing with the GPS
AI and Space: finally, no more arguing with the GPS
Speck&Tech171 visualizações
Space is open for business por Speck&Tech
Space is open for businessSpace is open for business
Space is open for business
Speck&Tech176 visualizações
Building large science space projects por Speck&Tech
Building large science space projectsBuilding large science space projects
Building large science space projects
Speck&Tech161 visualizações
Breaking the Laws of Robotics: Attacking Industrial Robots por Speck&Tech
Breaking the Laws of Robotics: Attacking Industrial RobotsBreaking the Laws of Robotics: Attacking Industrial Robots
Breaking the Laws of Robotics: Attacking Industrial Robots
Speck&Tech176 visualizações
Log Stealers - Shopping time for Threat Actors! por Speck&Tech
Log Stealers - Shopping time for Threat Actors!Log Stealers - Shopping time for Threat Actors!
Log Stealers - Shopping time for Threat Actors!
Speck&Tech368 visualizações
Design beyond deliverables por Speck&Tech
Design beyond deliverablesDesign beyond deliverables
Design beyond deliverables
Speck&Tech192 visualizações
Your website is ugly - how to fix it (and why you should care) por Speck&Tech
Your website is ugly - how to fix it (and why you should care)Your website is ugly - how to fix it (and why you should care)
Your website is ugly - how to fix it (and why you should care)
Speck&Tech215 visualizações
Why you should get a worse computer por Speck&Tech
Why you should get a worse computerWhy you should get a worse computer
Why you should get a worse computer
Speck&Tech228 visualizações
Intelligenza artificiale: uno yin o uno yang per i diritti umani? por Speck&Tech
Intelligenza artificiale: uno yin o uno yang per i diritti umani?Intelligenza artificiale: uno yin o uno yang per i diritti umani?
Intelligenza artificiale: uno yin o uno yang per i diritti umani?
Speck&Tech175 visualizações

Último

iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas... por
iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...
iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...Bernd Ruecker
33 visualizações69 slides
HTTP headers that make your website go faster - devs.gent November 2023 por
HTTP headers that make your website go faster - devs.gent November 2023HTTP headers that make your website go faster - devs.gent November 2023
HTTP headers that make your website go faster - devs.gent November 2023Thijs Feryn
21 visualizações151 slides
Evolving the Network Automation Journey from Python to Platforms por
Evolving the Network Automation Journey from Python to PlatformsEvolving the Network Automation Journey from Python to Platforms
Evolving the Network Automation Journey from Python to PlatformsNetwork Automation Forum
12 visualizações21 slides
PRODUCT PRESENTATION.pptx por
PRODUCT PRESENTATION.pptxPRODUCT PRESENTATION.pptx
PRODUCT PRESENTATION.pptxangelicacueva6
13 visualizações1 slide
Tunable Laser (1).pptx por
Tunable Laser (1).pptxTunable Laser (1).pptx
Tunable Laser (1).pptxHajira Mahmood
24 visualizações37 slides
STPI OctaNE CoE Brochure.pdf por
STPI OctaNE CoE Brochure.pdfSTPI OctaNE CoE Brochure.pdf
STPI OctaNE CoE Brochure.pdfmadhurjyapb
13 visualizações1 slide

Último(20)

iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas... por Bernd Ruecker
iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...
iSAQB Software Architecture Gathering 2023: How Process Orchestration Increas...
Bernd Ruecker33 visualizações
HTTP headers that make your website go faster - devs.gent November 2023 por Thijs Feryn
HTTP headers that make your website go faster - devs.gent November 2023HTTP headers that make your website go faster - devs.gent November 2023
HTTP headers that make your website go faster - devs.gent November 2023
Thijs Feryn21 visualizações
Evolving the Network Automation Journey from Python to Platforms por Network Automation Forum
Evolving the Network Automation Journey from Python to PlatformsEvolving the Network Automation Journey from Python to Platforms
Evolving the Network Automation Journey from Python to Platforms
Network Automation Forum12 visualizações
PRODUCT PRESENTATION.pptx por angelicacueva6
PRODUCT PRESENTATION.pptxPRODUCT PRESENTATION.pptx
PRODUCT PRESENTATION.pptx
angelicacueva613 visualizações
Tunable Laser (1).pptx por Hajira Mahmood
Tunable Laser (1).pptxTunable Laser (1).pptx
Tunable Laser (1).pptx
Hajira Mahmood24 visualizações
STPI OctaNE CoE Brochure.pdf por madhurjyapb
STPI OctaNE CoE Brochure.pdfSTPI OctaNE CoE Brochure.pdf
STPI OctaNE CoE Brochure.pdf
madhurjyapb13 visualizações
virtual reality.pptx por G036GaikwadSnehal
virtual reality.pptxvirtual reality.pptx
virtual reality.pptx
G036GaikwadSnehal11 visualizações
TouchLog: Finger Micro Gesture Recognition Using Photo-Reflective Sensors por sugiuralab
TouchLog: Finger Micro Gesture Recognition  Using Photo-Reflective SensorsTouchLog: Finger Micro Gesture Recognition  Using Photo-Reflective Sensors
TouchLog: Finger Micro Gesture Recognition Using Photo-Reflective Sensors
sugiuralab19 visualizações
Mini-Track: Challenges to Network Automation Adoption por Network Automation Forum
Mini-Track: Challenges to Network Automation AdoptionMini-Track: Challenges to Network Automation Adoption
Mini-Track: Challenges to Network Automation Adoption
Network Automation Forum12 visualizações
handbook for web 3 adoption.pdf por Liveplex
handbook for web 3 adoption.pdfhandbook for web 3 adoption.pdf
handbook for web 3 adoption.pdf
Liveplex22 visualizações
PharoJS - Zürich Smalltalk Group Meetup November 2023 por Noury Bouraqadi
PharoJS - Zürich Smalltalk Group Meetup November 2023PharoJS - Zürich Smalltalk Group Meetup November 2023
PharoJS - Zürich Smalltalk Group Meetup November 2023
Noury Bouraqadi126 visualizações
Business Analyst Series 2023 - Week 3 Session 5 por DianaGray10
Business Analyst Series 2023 -  Week 3 Session 5Business Analyst Series 2023 -  Week 3 Session 5
Business Analyst Series 2023 - Week 3 Session 5
DianaGray10237 visualizações
Piloting & Scaling Successfully With Microsoft Viva por Richard Harbridge
Piloting & Scaling Successfully With Microsoft VivaPiloting & Scaling Successfully With Microsoft Viva
Piloting & Scaling Successfully With Microsoft Viva
Richard Harbridge12 visualizações
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N... por James Anderson
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...
GDG Cloud Southlake 28 Brad Taylor and Shawn Augenstein Old Problems in the N...
James Anderson66 visualizações
PRODUCT LISTING.pptx por angelicacueva6
PRODUCT LISTING.pptxPRODUCT LISTING.pptx
PRODUCT LISTING.pptx
angelicacueva613 visualizações
Automating a World-Class Technology Conference; Behind the Scenes of CiscoLive por Network Automation Forum
Automating a World-Class Technology Conference; Behind the Scenes of CiscoLiveAutomating a World-Class Technology Conference; Behind the Scenes of CiscoLive
Automating a World-Class Technology Conference; Behind the Scenes of CiscoLive
Network Automation Forum30 visualizações
Case Study Copenhagen Energy and Business Central.pdf por Aitana
Case Study Copenhagen Energy and Business Central.pdfCase Study Copenhagen Energy and Business Central.pdf
Case Study Copenhagen Energy and Business Central.pdf
Aitana16 visualizações
Empathic Computing: Delivering the Potential of the Metaverse por Mark Billinghurst
Empathic Computing: Delivering  the Potential of the MetaverseEmpathic Computing: Delivering  the Potential of the Metaverse
Empathic Computing: Delivering the Potential of the Metaverse
Mark Billinghurst476 visualizações

Machine learning with quantum computers

  • 1. Machine Learning with Quantum Computers Speck&Tech event Davide Pastorello Department of Mathematics, University of Bologna and Trento Institute for Fundamental Physics and Applications, INFN Impact Hub Trento. September 21, 2023 Picture credit: IBM
  • 2. Premise: The World is quantum At microscopic length scales (molecules, atoms, electrons,...): Newton’s laws of dynamics and electromagnetism based on Maxwell’s equations do not work. We need quantum mechanics to describe physical phenomena. Quantum things: • Superposition of states (is the electron either "here" or "there"? In both places!). • Observations alters the state of the observed system. (We observe the electron "here" with some probability) • Entanglement ("spooky action at distance" as Einstein said) Using physical phenomena to process information is a good and old idea... What if are we able to use quantum phenomena?
  • 3. A bit of history... 1981. At the First Conference of Physics of Computation (MIT), R. Feynman observed that quantum systems cannot be efficiently simulated by classical computers and proposed a model of quantum computer. 1985. D. Deutsch described the universal quantum computer in terms of the quantum Turing machine. 1994. P. Shor proposed an algorithm for a quantum computer to factor large integers in polynomial time. 1995. L. Grover presents the quantum search algorithm in an unsorted database. 2000s. First prototypes of working quantum computers. 2010s. Quantum computers on the market (e.g. D-Wave, IBM Q System 1).
  • 4. Present days (2023) • Anyone can run his/her quantum algorithms on available quantum machines (a basic access is usually free). • Real-life problems are tackled by quantum computers. • The interest by industry and non academic institutions is dramatically increasing. • Availability of fault-tolerant, error-corrected, universal quantum computers? Not yet, we are in the Noisy Intermediate-Scale Quantum (NISQ). • Quantum supremacy? Not yet. • Hot topics: quantum optimization and quantum machine learning...
  • 5. Quantum computers exist Examples of quantum computers: D-Wave IBM Quantum
  • 6. Quantum bits Examples of qubits: • Particles with spin 1/2; • Polarized Photons; • Controlled superconducting circuits.
  • 7. Qubits (mathematical description) Quantum state superposition The state of a qubit is a unit vector* in C2: | i = ↵|0i + |1i ↵, 2 C , |↵|2 + | |2 = 1 {|0i, |1i} computational basis. A measurement process affects the qubit state Measurement of a qubit (e.g. we measure the polarization of a photon) P(0) = |↵|2 P(1) = | |2 The qubit state after the measurement is |ii if the outcome is i = 0, 1. *The vectors | i and ei✓ | i with ✓ 2 R represent the same physical state.
  • 8. Qubits (mathematical description) Tensor product | i = ✓ ↵ ◆ 2 C2 |'i = ✓ ◆ 2 C2 | i ⌦ |'i = 0 B B @ ↵ ↵ 1 C C A 2 C4 2 qubits (as a composite system) are described in: C2 ⌦ C2 := span{| i ⌦ |'i : | i, |'i 2 C2 } = C4 . n qubits (as a composite system) are described in: (C2 )⌦n = C2n .
  • 9. A new kind of information Empirical evidence: Quantum randomness, state superpositions, entanglement are physical phenomena not simply theoretical interpretations. + Encoding information into qubits (or more general quantum objects) allows completely new kinds of: • information storing and processing • data representation • telecommunication • cybersecurity • ... In particular we are interested in machine learning!
  • 10. Quantum gates Let {|0i, |1i} be the computational basis • Hadamard gate: |xi H 1 p 2 (|0i + ( 1)x |1i) x 2 {0, 1} • Phase gate: ↵|0i + |1i P ↵|0i + ei |1i • CNOT gate: |xi • |xi |yi |x yi x, y 2 {0, 1} Theorem: {H, P⇡/4, CNOT} is a universal set for quantum computation.
  • 11. Machine learning with quantum computers? Achievements in Quantum Machine Learning (since 2013) Algorithm Quantum speedup K-medians Quadratic Hierarchical clustering Quadratic K-means Exponential Principal component analysis Exponential Support vector machines Exponential Several classifiers Exponential Neural networks ? General observation Quantum architectures enable new paradigms of data representation and information processing. Current challenge Devising quantum learning mechanisms for available or near-term quantum machines.
  • 12. A quantum binary classifier The classification model Assume data are represented in a feature space. The model assigns a binary label to a new data instance evaluating its similarity to any training point. Quantum implementation on the IBM ibmq_melbourne • Quantum superposition of training vectors; • Test instance in quantum superposition of the two classes; • Parallel computing over the superposed data. Classical complexity: O(Nd) Quantum complexity: O(log(Nd)) Reference: D. Pastorello, E. Blanzieri. A quantum binary classifier based on cosine similarity Proc. IEEE International Conference on Quantum Computing and Engineering (QCE 21)
  • 13. Some QML schemes designed for universal quantum computers • Quantum divisive clustering W. Aïmeur et al. Quantum clustering algorithms ICML ’07: Proceedings of the 24th international conference on Machine learning (2007) • Quantum principal component analysis S. Lloyd, et al. Quantum principal component analysis Nature Physics 10, 631 (2014) • Quantum support vector machine P. Rebentrost et al. Quantum support vector machine for big data classification Phys. Rev. Lett. 113, 130503 (2014) • Quantum nearest neighbor N. Wiebe et al. Quantum Algorithms for Nearest-Neighbor Methods for Supervised and Unsupervised Learning Quantum Information and Computation 15(3,4): 0318- 0358 (2015) • Quantum perceptron M. Schuld et al. Simulating a perceptron on a quantum computer Physics Letters A, 379, pp. 660-663 (2015) • Quantum meta-learning M. Wilson et al. Optimizing quantum heuristic with meta-learning Quantum Machine Intelligence 3 (2021)
  • 14. Quantum Annealing Quantum Annealers The hardware is a quantum spin glass, i.e. a collection of qubits arranged in the vertices of a graph (V , E) where edges represent the interactions between neighbors. Example: D-Wave Chimera topology
  • 15. Quantum Annealing Annealing process (annealing time ⇠ 20µs) By energy dissipation the quantum system evolves in the ground state (the less energetic state) corresponding to the solution of a given optimization problem.
  • 16. Quantum Annealing Shortcoming Representing a given problem into the quantum hardware architecture is difficult in general and can destroy the efficiency of the quantum computation. Proposed solution Quantum Annealing Learning Search (QALS): A hybrid quantum-classical algorithm enabling a learning mechanism so that the quantum machine learns the representation of the problem on its own without an expensive classical pre-processing. References: -) Davide Pastorello, E. Blanzieri. Quantum Annealing Learning Search for solving QUBO problems. Quantum Information Processing 18: 303 (2019) -) A. Bonomi et al. Quantum annealing learning search implementations in Quantum Information and Computation, v. 22, n. 3&4 (2022)
  • 17. ML with quantum annealers • Boltzmann machine implementation M.H. Amin Quantum Boltzmann machine Phys. Rev. X 8, 021050 (2018) • Classification N. T. Nguyen et al. Image classification using quantum inference on the D-wave 2X. In: 2018 IEEE International Conference on Rebooting Computing (ICRC), pp. 1-7. IEEE (2018) • Clustering V. Kumar et al. Quantum Annealing for Combinatorial Clustering Quantum Inf Process (2018) 17: 39 • Training of a SVM D. Willsch et al. Support vector machines on the D-wave quantum annealer Computer Physics Communications 248, 107006 (2020)
  • 18. ...and quantum neural networks? A feedforward QNN made by single-qubit rotations and CNOTs: A QNN of this form achieves higher values of effective dimension w.r.t. to a classical NN with same number of parameters. Ref.: A. Abbas et al. The power of quantum neural networks Nat. Comput. Sci. 1, 403-409 (2021)
  • 19. Some concluding remarks • Quantum architectures enable new kinds of data processing. • There is a bottleneck about the efficient data representation into quantum hardware (QRAM are not feasible yet). • The era of quantum machine learning is just started but it is maybe the most promising road for putting forward the frontiers of quantum computing. • The quantum-classical hybrid approach is the best way to use the existing quantum machines. • Providing efficient quantum algorithms is a very difficult task. So we have to devise learning mechanisms of new quantum algorithms towards a "quantum meta-learning"... In other words, quantum computers must learn how to solve problems on their own...
  • 20. Thanks for your attention! Picture credit: IBM