Papers
Topics
Authors
Recent
2000 character limit reached

Approximately-symmetric neural networks for quantum spin liquids (2405.17541v1)

Published 27 May 2024 in quant-ph, cond-mat.dis-nn, cond-mat.str-el, and cs.LG

Abstract: We propose and analyze a family of approximately-symmetric neural networks for quantum spin liquid problems. These tailored architectures are parameter-efficient, scalable, and significantly out-perform existing symmetry-unaware neural network architectures. Utilizing the mixed-field toric code model, we demonstrate that our approach is competitive with the state-of-the-art tensor network and quantum Monte Carlo methods. Moreover, at the largest system sizes (N=480), our method allows us to explore Hamiltonians with sign problems beyond the reach of both quantum Monte Carlo and finite-size matrix-product states. The network comprises an exactly symmetric block following a non-symmetric block, which we argue learns a transformation of the ground state analogous to quasiadiabatic continuation. Our work paves the way toward investigating quantum spin liquid problems within interpretable neural network architectures

Definition Search Book Streamline Icon: https://streamlinehq.com
References (89)
  1. P. W. Anderson, Resonating valence bonds: A new kind of insulator?, Materials Research Bulletin 8, 153 (1973).
  2. A. Kitaev, Anyons in an exactly solved model and beyond, Annals of Physics 321, 2 (2006).
  3. L. Savary and L. Balents, Quantum spin liquids: a review, Reports on Progress in Physics 80, 016502 (2016).
  4. R. Verresen, M. D. Lukin, and A. Vishwanath, Prediction of toric code topological order from rydberg blockade, Physical Review X 11, 031005 (2021).
  5. Non-abelian braiding of graph vertices in a superconducting processor, Nature 618, 264 (2023).
  6. Please see the Supplementary Materials.
  7. M. B. Hastings and X.-G. Wen, Quasiadiabatic continuation of quantum states: The stability of topological ground-state degeneracy and emergent gauge invariance, Physical review b 72, 045141 (2005).
  8. U. Schollwöck, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326, 96 (2011), january 2011 Special Issue.
  9. M. C. Bañuls, Tensor network algorithms: A route map, Annual Review of Condensed Matter Physics 14, 173 (2023).
  10. W. L. McMillan, Ground state of liquid he4superscripthe4{\mathrm{he}}^{4}roman_he start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT, Phys. Rev. 138, A442 (1965).
  11. D. Ceperley, G. V. Chester, and M. H. Kalos, Monte carlo simulation of a many-fermion study, Phys. Rev. B 16, 3081 (1977).
  12. P. R. C. Kent, R. J. Needs, and G. Rajagopal, Monte carlo energy and variance-minimization techniques for optimizing many-body wave functions, Phys. Rev. B 59, 12344 (1999).
  13. G. Carleo and M. Troyer, Solving the quantum many-body problem with artificial neural networks, Science 355, 602 (2017).
  14. K. Hornik, M. Stinchcombe, and H. White, Multilayer feedforward networks are universal approximators, Neural networks 2, 359 (1989).
  15. O. Sharir, A. Shashua, and G. Carleo, Neural tensor contractions and the expressive power of deep neural quantum states, Physical Review B 106, 205136 (2022).
  16. K. Sprague and S. Czischek, Variational monte carlo with large patched transformers, arXiv preprint arXiv:2306.03921  (2023).
  17. P. Patil, Quantum monte carlo simulations in the restricted hilbert space of rydberg atom arrays, arXiv preprint arXiv:2309.00482  (2023).
  18. L. L. Viteritti, F. Ferrari, and F. Becca, Accuracy of restricted boltzmann machines for the one-dimensional j⁢_⁢1−j⁢_⁢2𝑗_1𝑗_2j\_1-j\_2italic_j _ 1 - italic_j _ 2 heisenberg model, SciPost Physics 12, 166 (2022).
  19. K. Choo, T. Neupert, and G. Carleo, Two-dimensional frustrated j 1- j 2 model studied with neural network quantum states, Physical Review B 100, 125124 (2019).
  20. C. Roth, A. Szabó, and A. H. MacDonald, High-accuracy variational monte carlo for frustrated magnets with deep neural networks, Physical Review B 108, 054410 (2023).
  21. T. Cohen and M. Welling, Group equivariant convolutional networks, in International conference on machine learning (PMLR, 2016) pp. 2990–2999.
  22. M. Reh, M. Schmitt, and M. Gärttner, Optimizing design choices for neural quantum states, Physical Review B 107, 195115 (2023).
  23. M. Finzi, G. Benton, and A. G. Wilson, Residual pathway priors for soft equivariance constraints, Advances in Neural Information Processing Systems 34, 30037 (2021).
  24. R. Wang, R. Walters, and R. Yu, Approximately equivariant networks for imperfectly symmetric dynamics, in International Conference on Machine Learning (PMLR, 2022) pp. 23078–23091.
  25. F. Wu, Y. Deng, and N. Prokof’ev, Phase diagram of the toric code model in a parallel magnetic field, Physical Review B 85, 195104 (2012).
  26. E. Fradkin and S. H. Shenker, Phase diagrams of lattice gauge theories with higgs fields, Phys. Rev. D 19, 3682 (1979).
  27. S. Sorella, Green function monte carlo with stochastic reconfiguration, Physical review letters 80, 4558 (1998).
  28. S.-I. Amari, Natural gradient works efficiently in learning, Neural computation 10, 251 (1998).
  29. Here the expectation 𝔼ssubscript𝔼𝑠\mathbb{E}_{s}blackboard_E start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT is calculated with respect to the probability p⁢(s)=|ψ⁢(s)|2𝑝𝑠superscript𝜓𝑠2p(s)=|\psi(s)|^{2}italic_p ( italic_s ) = | italic_ψ ( italic_s ) | start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT.
  30. S. R. White, Density matrix formulation for quantum renormalization groups, Phys. Rev. Lett. 69, 2863 (1992).
  31. S. R. White, Density-matrix algorithms for quantum renormalization groups, Phys. Rev. B 48, 10345 (1993).
  32. U. Schollwöck, The density-matrix renormalization group, Rev. Mod. Phys. 77, 259 (2005).
  33. J. Greitemann and L. Pollet, Lecture notes on diagrammatic monte carlo for the fröhlich polaron, SciPost Physics Lecture Notes , 002 (2018).
  34. J. c. v. Chaloupka, G. Jackeli, and G. Khaliullin, Kitaev-heisenberg model on a honeycomb lattice: Possible exotic phases in iridium oxides A2⁢iro3subscript𝐴2subscriptiro3{A}_{2}{\mathrm{iro}}_{3}italic_A start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT roman_iro start_POSTSUBSCRIPT 3 end_POSTSUBSCRIPT, Phys. Rev. Lett. 105, 027204 (2010).
  35. H.-C. Jiang, H. Yao, and L. Balents, Spin liquid ground state of the spin-1212\frac{1}{2}divide start_ARG 1 end_ARG start_ARG 2 end_ARG square J1subscript𝐽1{J}_{1}italic_J start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT-J2subscript𝐽2{J}_{2}italic_J start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT heisenberg model, Phys. Rev. B 86, 024424 (2012).
  36. J. Bricmont and J. Frölich, An order parameter distinguishing between different phases of lattice gauge theories with matter fields, Physics Letters B 122, 73 (1983).
  37. K. Fredenhagen and M. Marcu, Charged states in ℤℤ\mathbb{Z}blackboard_Z2 gauge theories, Communications in Mathematical Physics 92, 81 (1983).
  38. Note the factor of 1/2121/21 / 2 rescaling in the definition of magnetic fields between our work and [61].
  39. N. Diaconu and D. Worrall, Learning to convolve: A generalized weight-tying approach, in International Conference on Machine Learning (PMLR, 2019) pp. 1586–1595.
  40. T. Vieijra and J. Nys, Many-body quantum states with exact conservation of non-abelian and lattice symmetries through variational monte carlo, Physical Review B 104, 045123 (2021).
  41. Tom Vieijra, Corneel Casert, Jannes Nys, Wesley De Neve, Jutho Haegeman, Jan Ryckebusch,  and Frank Verstraete, “Restricted boltzmann machines for quantum states with non-abelian or anyonic symmetries,” Physical review letters 124, 097201 (2020).
  42. Kenny Choo, Titus Neupert,  and Giuseppe Carleo, “Two-dimensional frustrated j 1- j 2 model studied with neural network quantum states,” Physical Review B 100, 125124 (2019).
  43. Moritz Reh, Markus Schmitt,  and Martin Gärttner, “Optimizing design choices for neural quantum states,” Physical Review B 107, 195115 (2023).
  44. Zhuo Chen, Laker Newhouse, Eddie Chen, Di Luo,  and Marin Soljacic, “Antn: Bridging autoregressive neural networks and tensor networks for quantum many-body simulation,” Advances in Neural Information Processing Systems 36, 450–476 (2023).
  45. Di Luo, Shunyue Yuan, James Stokes,  and Bryan Clark, “Gauge equivariant neural networks for 2+ 1d u (1) gauge theory simulations in hamiltonian formulation,” in NeurIPS 2022 AI for Science: Progress and Promises (2022).
  46. Mohamed Hibat-Allah, Martin Ganahl, Lauren E Hayward, Roger G Melko,  and Juan Carrasquilla, “Recurrent neural network wave functions,” Physical Review Research 2, 023358 (2020).
  47. Taco Cohen and Max Welling, “Group equivariant convolutional networks,” in International conference on machine learning (PMLR, 2016) pp. 2990–2999.
  48. John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, et al., “Highly accurate protein structure prediction with alphafold,” Nature 596, 583–589 (2021).
  49. Christopher Roth, Attila Szabó,  and Allan H MacDonald, “High-accuracy variational monte carlo for frustrated magnets with deep neural networks,” Physical Review B 108, 054410 (2023).
  50. Extending the applicability of the method beyond bit string to bit string mapping is perhaps possible by proceeding along lines of Ref. [49] where one imposes S⁢U⁢(2)𝑆𝑈2SU(2)italic_S italic_U ( 2 ) + lattice symmetries within the constrained neural network architecture utilizing idea of an equivariant Clebsch-Gordan nets [50].
  51. University of Amsterdam QUVA lab, “Group-equivariant neural network demonstration,” https://github.com/dom-kufel/g_equiv_networks/blob/main/conventional_cnn.gif.
  52. Kenny Choo, Giuseppe Carleo, Nicolas Regnault,  and Titus Neupert, “Symmetries and many-body excitations with neural-network quantum states,” Physical review letters 121, 167204 (2018).
  53. Sékou-Oumar Kaba, Arnab Kumar Mondal, Yan Zhang, Yoshua Bengio,  and Siamak Ravanbakhsh, “Equivariance with learned canonicalization functions,” in International Conference on Machine Learning (PMLR, 2023) pp. 15546–15566.
  54. Sékou-Oumar Kaba and Siamak Ravanbakhsh, “Symmetry breaking and equivariant neural networks,” arXiv preprint arXiv:2312.09016  (2023).
  55. Marc Finzi, Max Welling,  and Andrew Gordon Wilson, “A practical method for constructing equivariant multilayer perceptrons for arbitrary matrix groups,” in International conference on machine learning (PMLR, 2021) pp. 3318–3328.
  56. Taco S Cohen and Max Welling, “Steerable cnns,” arXiv preprint arXiv:1612.08498  (2016b).
  57. Manzil Zaheer, Satwik Kottur, Siamak Ravanbakhsh, Barnabas Poczos, Russ R Salakhutdinov,  and Alexander J Smola, “Deep sets,” Advances in neural information processing systems 30 (2017).
  58. Maurice Weiler and Gabriele Cesa, “General e (2)-equivariant steerable cnns,” Advances in neural information processing systems 32 (2019).
  59. Maurice Weiler, Mario Geiger, Max Welling, Wouter Boomsma,  and Taco S Cohen, “3d steerable cnns: Learning rotationally equivariant features in volumetric data,” Advances in Neural Information Processing Systems 31 (2018).
  60. Agnes Valenti, Eliska Greplova, Netanel H Lindner,  and Sebastian D Huber, “Correlation-enhanced neural networks as interpretable variational quantum states,” Physical Review Research 4, L012010 (2022).
  61. Rui Wang, Robin Walters,  and Rose Yu, “Approximately equivariant networks for imperfectly symmetric dynamics,” in International Conference on Machine Learning (PMLR, 2022) pp. 23078–23091.
  62. Marc Finzi, Gregory Benton,  and Andrew G Wilson, “Residual pathway priors for soft equivariance constraints,” Advances in Neural Information Processing Systems 34, 30037–30049 (2021b).
  63. Kaiming He, Xiangyu Zhang, Shaoqing Ren,  and Jian Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition (2016) pp. 770–778.
  64. S Unnikrishna Pillai, Torsten Suel,  and Seunghun Cha, “The perron-frobenius theorem: some of its applications,” IEEE Signal Processing Magazine 22, 62–75 (2005).
  65. Marc Machaczek, “Neural quantum states for fracton models,” https://zenodo.org/records/10728168 (2024).
  66. Djork-Arné Clevert, Thomas Unterthiner,  and Sepp Hochreiter, “Fast and accurate deep network learning by exponential linear units (elus),” arXiv preprint arXiv:1511.07289  (2015).
  67. Sandro Sorella, “Green function monte carlo with stochastic reconfiguration,” Physical review letters 80, 4558 (1998).
  68. James Stokes, Josh Izaac, Nathan Killoran,  and Giuseppe Carleo, “Quantum natural gradient,” Quantum 4, 269 (2020).
  69. Giuseppe Carleo and Matthias Troyer, “Solving the quantum many-body problem with artificial neural networks,” Science 355, 602–606 (2017).
  70. Ao Chen and Markus Heyl, “Efficient optimization of deep neural quantum states toward machine precision,” arXiv preprint arXiv:2302.01941  (2023).
  71. Filippo Vicentini, Damian Hofmann, Attila Szabó, Dian Wu, Christopher Roth, Clemens Giuliani, Gabriel Pescia, Jannes Nys, Vladimir Vargas-Calderón, Nikita Astrakhantsev, et al., “Netket 3: Machine learning toolbox for many-body quantum systems,” SciPost Physics Codebases , 007 (2022).
  72. James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne,  and Qiao Zhang, “JAX: composable transformations of Python+NumPy programs,”  (2018).
  73. Jonathan Heek, Anselm Levskaya, Avital Oliver, Marvin Ritter, Bertrand Rondepierre, Andreas Steiner,  and Marc van Zee, “Flax: A neural network library and ecosystem for JAX,”  (2023).
  74. Matthew Fishman, Steven R. White,  and E. Miles Stoudenmire, “The ITensor Software Library for Tensor Network Calculations,” SciPost Phys. Codebases , 4 (2022a).
  75. Matthew Fishman, Steven R. White,  and E. Miles Stoudenmire, “Codebase release 0.3 for ITensor,” SciPost Phys. Codebases , 4–r0.3 (2022b).
  76. Gregory D. Kahanamoku-Meyer and Julia Wei, “GregDMeyer/dynamite: v0.4.0,”  (2024).
  77. Attila Szabó and Claudio Castelnovo, “Neural network wave functions and the sign problem,” Physical Review Research 2, 033075 (2020).
  78. Li Jing, Yichen Shen, Tena Dubcek, John Peurifoy, Scott Skirlo, Yann LeCun, Max Tegmark,  and Marin Soljačić, “Tunable efficient unitary neural networks (eunn) and their application to rnns,” in International Conference on Machine Learning (PMLR, 2017) pp. 1733–1741.
  79. Nikita Astrakhantsev, Tom Westerhout, Apoorv Tiwari, Kenny Choo, Ao Chen, Mark H Fischer, Giuseppe Carleo,  and Titus Neupert, “Broken-symmetry ground states of the heisenberg model on the pyrochlore lattice,” Physical Review X 11, 041021 (2021).
  80. Remmy Zen, Long My, Ryan Tan, Frédéric Hébert, Mario Gattobigio, Christian Miniatura, Dario Poletti,  and Stéphane Bressan, “Transfer learning for scalability of neural-network quantum states,” Physical Review E 101, 053301 (2020).
  81. Sergey Bravyi, David Gosset,  and Yinchen Liu, “How to simulate quantum measurement without computing marginals,” Physical Review Letters 128, 220503 (2022).
  82. Sergey Bravyi, Giuseppe Carleo, David Gosset,  and Yinchen Liu, “A rapidly mixing markov chain from any gapped quantum many-body system,” Quantum 7, 1173 (2023).
  83. Aki Vehtari, Andrew Gelman, Daniel Simpson, Bob Carpenter,  and Paul-Christian Bürkner, “Rank-normalization, folding, and localization: An improved r for assessing convergence of mcmc (with discussion),” Bayesian analysis 16, 667–718 (2021).
  84. Wen-Tao Xu, Frank Pollmann,  and Michael Knap, “Critical behavior of the fredenhagen-marcu order parameter at topological phase transitions,” arXiv preprint arXiv:2402.00127  (2024).
  85. Matthew B Hastings, Iván González, Ann B Kallin,  and Roger G Melko, “Measuring renyi entanglement entropy in quantum monte carlo simulations,” Physical review letters 104, 157201 (2010).
  86. Lack of symmetry between X𝑋Xitalic_X and Z𝑍Zitalic_Z perturbations comes from the open boundary conditions: it might be seen in perturbation theory that X𝑋Xitalic_X and Z𝑍Zitalic_Z magnetic field excitations at the boundary give contributions to energy in different orders of perturbation theory.
  87. Marc Finzi, Understanding and Incorporating Mathematical Inductive Biases in Neural Networks, Ph.D. thesis, New York University (2023).
  88. Tom Vieijra and Jannes Nys, “Many-body quantum states with exact conservation of non-abelian and lattice symmetries through variational monte carlo,” Physical Review B 104, 045123 (2021).
  89. Risi Kondor, Zhen Lin,  and Shubhendu Trivedi, “Clebsch–gordan nets: a fully fourier space spherical convolutional neural network,” Advances in Neural Information Processing Systems 31 (2018).
Citations (2)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
Slide Deck Streamline Icon: https://streamlinehq.com

Whiteboard

Dice Question Streamline Icon: https://streamlinehq.com

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Generate Now
Lightbulb Streamline Icon: https://streamlinehq.com

Continue Learning

We haven't generated follow-up questions for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Generate Now
List To Do Tasks Checklist Streamline Icon: https://streamlinehq.com

Collections

Sign up for free to add this paper to one or more collections.

User Circle Single Streamline Icon: https://streamlinehq.com Sign Up
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets

This paper has been mentioned in 2 tweets and received 0 likes.

Upgrade to Pro to view all of the tweets about this paper:

Start a free 7-day Pro trial