Learning about learning by many-body systems

Abstract

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.

arXiv comment10 pages (with 7 figures) + appendices. Close to published version. This version, including the appendices, subsumes arXiv:2001.03623

References in corpus (19)

• Natural Scales in Geographical Patterns
• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Machine learning and the physical sciences
• Spontaneous motion in hierarchically assembled active matter
• Tutorial on Variational Autoencoders
• Discovering physical concepts with neural networks
• Memory formation in matter
• Multiple transient memories in experiments on sheared non-Brownian suspensions
• Generic transient memory formation in disordered systems with noise
• Representation Learning: A Review and New Perspectives
• Toward an AI Physicist for Unsupervised Learning
• Denoising Criterion for Variational Auto-Encoding Framework
• Strength of Mechanical Memories is Maximal at the Yield Point of a Soft Glass
• Multiple transient memories in sheared suspensions: robustness, structure, and routes to plasticity
• Associative pattern recognition through macro-molecular self-assembly
• Self-Organized Short-Term Memories
• TherML: Thermodynamics of Machine Learning
• Noise Stabilization of Self-Organized Memories
• Self-organized novelty detection in driven spin glasses

Cited by in corpus (4)

• Thermodynamic Computing via Autonomous Quantum Thermal Machines
• From observations to complexity of quantum states via unsupervised learning
• Experimentally measuring rolling and sliding in three-dimensional dense granular packings
• Memory in 3D cyclically driven granular material