Representation Learning via Quantum Neural Tangent Kernels
Abstract
Variational quantum circuits are used in quantum machine learning and variational quantum simulation tasks. Designing good variational circuits or predicting how well they perform for given learning or optimization tasks is still unclear. Here we discuss these problems, analyzing variational quantum circuits using the theory of neural tangent kernels. We define quantum neural tangent kernels, and derive dynamical equations for their associated loss function in optimization and learning tasks. We analytically solve the dynamics in the frozen limit, or lazy training regime, where variational angles change slowly and a linear perturbation is good enough. We extend the analysis to a dynamical setting, including quadratic corrections in the variational angles. We then consider a hybrid quantum classical architecture and define a large-width limit for hybrid kernels, showing that a hybrid quantum classical neural network can be approximately Gaussian. The results presented here show limits for which analytical understandings of the training dynamics for variational quantum circuits, used for quantum machine learning and optimization problems, are possible. These analytical results are supported by numerical simulations of quantum machine-learning experiments.
- Authors:
-
- Univ. of Chicago, IL (United States); Chicago Quantum Exchange, IL (United States)
- IBM Research-Zurich, Rüschlikon (Switzerland)
- IBM, Yorktown Heights, NY (United States). Thomas J. Watson Research Center
- Publication Date:
- Research Org.:
- National Quantum Information Science (QIS) Research Centers (United States). Next Generation Quantum Science and Engineering (Q-NEXT)
- Sponsoring Org.:
- USDOE; US Air Force Office of Scientific Research (AFOSR); US Army Research Office (ARO); National Science Foundation (NSF)
- OSTI Identifier:
- 1982853
- Grant/Contract Number:
- AC02-06CH11357; FA9550-21-1-0209; W911NF-18-1-0020; W911NF-18-1-0212; W911NF-16-1-0349; FA9550-19-1-0399; EFMA-1640959; OMA-1936118; EEC-1941583
- Resource Type:
- Accepted Manuscript
- Journal Name:
- PRX Quantum
- Additional Journal Information:
- Journal Volume: 3; Journal Issue: 3; Journal ID: ISSN 2691-3399
- Publisher:
- American Physical Society (APS)
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; Physics; Machine learning; Quantum algorithms; Quantum simulation
Citation Formats
Liu, Junyu, Tacchino, Francesco, Glick, Jennifer R., Jiang, Liang, and Mezzacapo, Antonio. Representation Learning via Quantum Neural Tangent Kernels. United States: N. p., 2022.
Web. doi:10.1103/prxquantum.3.030323.
Liu, Junyu, Tacchino, Francesco, Glick, Jennifer R., Jiang, Liang, & Mezzacapo, Antonio. Representation Learning via Quantum Neural Tangent Kernels. United States. https://doi.org/10.1103/prxquantum.3.030323
Liu, Junyu, Tacchino, Francesco, Glick, Jennifer R., Jiang, Liang, and Mezzacapo, Antonio. Wed .
"Representation Learning via Quantum Neural Tangent Kernels". United States. https://doi.org/10.1103/prxquantum.3.030323. https://www.osti.gov/servlets/purl/1982853.
@article{osti_1982853,
title = {Representation Learning via Quantum Neural Tangent Kernels},
author = {Liu, Junyu and Tacchino, Francesco and Glick, Jennifer R. and Jiang, Liang and Mezzacapo, Antonio},
abstractNote = {Variational quantum circuits are used in quantum machine learning and variational quantum simulation tasks. Designing good variational circuits or predicting how well they perform for given learning or optimization tasks is still unclear. Here we discuss these problems, analyzing variational quantum circuits using the theory of neural tangent kernels. We define quantum neural tangent kernels, and derive dynamical equations for their associated loss function in optimization and learning tasks. We analytically solve the dynamics in the frozen limit, or lazy training regime, where variational angles change slowly and a linear perturbation is good enough. We extend the analysis to a dynamical setting, including quadratic corrections in the variational angles. We then consider a hybrid quantum classical architecture and define a large-width limit for hybrid kernels, showing that a hybrid quantum classical neural network can be approximately Gaussian. The results presented here show limits for which analytical understandings of the training dynamics for variational quantum circuits, used for quantum machine learning and optimization problems, are possible. These analytical results are supported by numerical simulations of quantum machine-learning experiments.},
doi = {10.1103/prxquantum.3.030323},
journal = {PRX Quantum},
number = 3,
volume = 3,
place = {United States},
year = {Wed Aug 17 00:00:00 EDT 2022},
month = {Wed Aug 17 00:00:00 EDT 2022}
}
Works referenced in this record:
Quantum Support Vector Machine for Big Data Classification
journal, September 2014
- Rebentrost, Patrick; Mohseni, Masoud; Lloyd, Seth
- Physical Review Letters, Vol. 113, Issue 13
Quantum computational chemistry
journal, March 2020
- McArdle, Sam; Endo, Suguru; Aspuru-Guzik, Alán
- Reviews of Modern Physics, Vol. 92, Issue 1
Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks
journal, May 2021
- Canatar, Abdulkadir; Bordelon, Blake; Pehlevan, Cengiz
- Nature Communications, Vol. 12, Issue 1
Quantum Machine Learning in Feature Hilbert Spaces
journal, February 2019
- Schuld, Maria; Killoran, Nathan
- Physical Review Letters, Vol. 122, Issue 4
Information-Theoretic Bounds on Quantum Advantage in Machine Learning
journal, May 2021
- Huang, Hsin-Yuan; Kueng, Richard; Preskill, John
- Physical Review Letters, Vol. 126, Issue 19
Quantum Algorithm for Linear Systems of Equations
journal, October 2009
- Harrow, Aram W.; Hassidim, Avinatan; Lloyd, Seth
- Physical Review Letters, Vol. 103, Issue 15
Chaos and complexity by design
journal, April 2017
- Roberts, Daniel A.; Yoshida, Beni
- Journal of High Energy Physics, Vol. 2017, Issue 4
Variational quantum algorithms
journal, August 2021
- Cerezo, M.; Arrasmith, Andrew; Babbush, Ryan
- Nature Reviews Physics, Vol. 3, Issue 9
Supervised learning with quantum-enhanced feature spaces
journal, March 2019
- Havlíček, Vojtěch; Córcoles, Antonio D.; Temme, Kristan
- Nature, Vol. 567, Issue 7747
Barren plateaus in quantum neural network training landscapes
journal, November 2018
- McClean, Jarrod R.; Boixo, Sergio; Smelyanskiy, Vadim N.
- Nature Communications, Vol. 9, Issue 1
Neural networks and quantum field theory
journal, April 2021
- Halverson, James; Maiti, Anindita; Stoner, Keegan
- Machine Learning: Science and Technology, Vol. 2, Issue 3
Quantum Algorithm for Data Fitting
journal, August 2012
- Wiebe, Nathan; Braun, Daniel; Lloyd, Seth
- Physical Review Letters, Vol. 109, Issue 5
Scrambling and decoding the charged quantum information
journal, October 2020
- Liu, Junyu
- Physical Review Research, Vol. 2, Issue 4
Anomalous collapses of Nares Strait ice arches leads to enhanced export of Arctic sea ice
journal, January 2021
- Moore, G. W. K.; Howell, S. E. L.; Brady, M.
- Nature Communications, Vol. 12, Issue 1
Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets
journal, September 2017
- Kandala, Abhinav; Mezzacapo, Antonio; Temme, Kristan
- Nature, Vol. 549, Issue 7671
Quantum principal component analysis
journal, July 2014
- Lloyd, Seth; Mohseni, Masoud; Rebentrost, Patrick
- Nature Physics, Vol. 10, Issue 9
The theory of variational hybrid quantum-classical algorithms
journal, February 2016
- McClean, Jarrod R.; Romero, Jonathan; Babbush, Ryan
- New Journal of Physics, Vol. 18, Issue 2
A variational eigenvalue solver on a photonic quantum processor
journal, July 2014
- Peruzzo, Alberto; McClean, Jarrod; Shadbolt, Peter
- Nature Communications, Vol. 5, Issue 1
The power of quantum neural networks
journal, June 2021
- Abbas, Amira; Sutter, David; Zoufal, Christa
- Nature Computational Science, Vol. 1, Issue 6
A rigorous and robust quantum speed-up in supervised machine learning
journal, July 2021
- Liu, Yunchao; Arunachalam, Srinivasan; Temme, Kristan
- Nature Physics, Vol. 17, Issue 9
Quantum machine learning
journal, September 2017
- Biamonte, Jacob; Wittek, Peter; Pancotti, Nicola
- Nature, Vol. 549, Issue 7671