A Caltech Library Service

Machine learning outperforms thermodynamics in measuring how well a many-body system learns a drive

Zhong, Weishun and Gold, Jacob M. and Marzen, Sarah and England, Jeremy L. and Yunger Halpern, Nicole (2021) Machine learning outperforms thermodynamics in measuring how well a many-body system learns a drive. Scientific Reports, 11 . Art. No. 9333. ISSN 2045-2322. doi:10.1038/s41598-021-88311-7.

[img] PDF - Published Version
Creative Commons Attribution.

[img] PDF - Supplemental Material
Creative Commons Attribution.


Use this Persistent URL to link to this item:


Diverse many-body systems, from soap bubbles to suspensions to polymers, learn and remember patterns in the drives that push them far from equilibrium. This learning may be leveraged for computation, memory, and engineering. Until now, many-body learning has been detected with thermodynamic properties, such as work absorption and strain. We progress beyond these macroscopic properties first defined for equilibrium contexts: We quantify statistical mechanical learning using representation learning, a machine-learning model in which information squeezes through a bottleneck. By calculating properties of the bottleneck, we measure four facets of many-body systems’ learning: classification ability, memory capacity, discrimination ability, and novelty detection. Numerical simulations of a classical spin glass illustrate our technique. This toolkit exposes self-organization that eludes detection by thermodynamic measures: Our toolkit more reliably and more precisely detects and quantifies learning by matter while providing a unifying framework for many-body learning.

Item Type:Article
Related URLs:
URLURL TypeDescription ItemCode
Marzen, Sarah0000-0001-5386-1101
Yunger Halpern, Nicole0000-0001-8670-6212
Additional Information:© The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit Received 24 November 2020. Accepted 01 April 2021. Published 29 April 2021. The authors thank Alexander Alemi, Isaac Chuang, Emine Kucukbenli, Nick Litombe, Seth Lloyd, Julia Steinberg, Tailin Wu, and Susanne Yelin for useful discussions. WZ is supported by ARO Grant W911NF-18-1-0101; the Gordon and Betty Moore Foundation Grant, under No. GBMF4343; and the Henry W. Kendall (1955) Fellowship Fund. JMG is funded by the AFOSR, under Grant FA9950-17-1-0136. SM was supported partially by the Moore Foundation, via the Physics of Living Systems Fellowship. This material is based upon work supported by, or in part by, the Air Force Office of Scientific Research, under award number FA9550-19-1-0411. JLE has been funded by the Air Force Office of Scientific Research Grant FA9550-17-1-0136 and by the James S. McDonnell Foundation Scholar Grant 220020476. NYH is grateful for an NSF grant for the Institute for Theoretical Atomic, Molecular, and Optical Physics at Harvard University and the Smithsonian Astrophysical Observatory. NYH also thanks CQIQC at the University of Toronto, the Fields Institute, and Caltech’s Institute for Quantum Information and Matter (NSF Grant PHY-1733907) for their hospitality during the development of this paper. These authors contributed equally: Weishun Zhong and Jacob M. Gold. Author Contributions. J.M.G. simulated the spin glass. W.Z. built the machine-learning code and processed the spin-glass data. S.M. built the novelty-detection code. N.Y.H. managed the project and wrote the manuscript. All authors (W.Z., J.M.G., S.M., J.L.E., and N.Y.H.) contributed to the project design and data analysis. W.Z. and J.M.G. contributed equally. Data availability. The machine-learning and spin-glass-simulation code is available at Ref.20. Will be available once COVID-19 restrictions loosen enough that we can access the computers that store the files. The authors declare no competing interests.
Group:Institute for Quantum Information and Matter
Funding AgencyGrant Number
Army Research Office (ARO)W911NF-18-1-0101
Gordon and Betty Moore FoundationGBMF4343
Henry W. Kendall (1955) Fellowship FundUNSPECIFIED
Air Force Office of Scientific Research (AFOSR)FA9950-17-1-0136
Air Force Office of Scientific Research (AFOSR)FA9550-19-1-0411
James S. McDonnell Foundation220020476
Record Number:CaltechAUTHORS:20210429-144552683
Persistent URL:
Official Citation:Zhong, W., Gold, J.M., Marzen, S. et al. Machine learning outperforms thermodynamics in measuring how well a many-body system learns a drive. Sci Rep 11, 9333 (2021).
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:108878
Deposited By: George Porter
Deposited On:30 Apr 2021 14:03
Last Modified:16 Nov 2021 19:33

Repository Staff Only: item control page