Demonstration of Decentralized, Physics-Driven Learning

Abstract

In typical artificial neural networks, neurons adjust according to global calculations of a central processor, but in the brain neurons and synapses self-adjust based on local information. Contrastive learning algorithms have recently been proposed to train physical systems, such as fluidic, mechanical, or electrical networks, to perform machine learning tasks from local evolution rules. However, to date such systems have only been implemented in silico due to the engineering challenge of creating elements that autonomously evolve based on their own response to two sets of global boundary conditions. Here we introduce and implement a physics-driven contrastive learning scheme for a network of variable resistors, using circuitry to locally compare the response of two identical networks subjected to the two different sets of boundary conditions. Using this innovation, our system effectively trains itself, optimizing its resistance values without use of a central processor or external information storage. Once the system is trained for a specified allostery, regression, or classification task, the task is subsequently performed rapidly and automatically by the physical imperative to minimize power dissipation in response to the given voltage inputs. We demonstrate that, unlike typical computers, such learning systems are robust to extreme damage (and thus manufacturing defects) due to their decentralized learning. Our twin-network approach is therefore readily scalable to extremely large or nonlinear networks where its distributed nature will be an enormous advantage; a laboratory network of only 500 edges will already outpace its in silico counterpart.

arXiv comment11 pages 7 figures

References in corpus (21)

• Overcoming catastrophic forgetting in neural networks
• Deep physical neural networks enabled by a backpropagation algorithm for arbitrary physical systems
• Towards Biologically Plausible Deep Learning
• Designing allostery-inspired response in mechanical networks
• Network Plasticity as Bayesian Inference
• Directed aging, memory and Nature's greed
• The limits of multifunctionality in tunable networks
• Supervised learning in a mechanical system
• Supervised learning in physical networks: From machine learning to learning machines
• Periodic training of creeping solids
• Training End-to-End Analog Neural Networks with Equilibrium Propagation
• Shaping dynamical folding and misfolding pathways in mechanical metamaterials
• Effect of aging on the non-linear elasticity and memory formation in materials
• The hidden topological structure of flow network functionality
• Desynchronous Learning in a Physics-Driven Learning Network
• Revealing structure-function relationships in functional flow networks via persistent homology
• Learned multi-stability in mechanical networks
• Local rules for fabricating allosteric networks
• Early Inference in Energy-Based Models Approximates Back-Propagation
• Tuning and jamming reduced to their minima
• Physical learning beyond the quasistatic limit

Cited by in corpus (21)

• Learning without neurons in physical systems
• Soft Metamaterials: Adaptation and Intelligence
• Machine Learning Without a Processor: Emergent Learning in a Nonlinear Electronic Metamaterial
• The Physical Effects of Learning
• Learning to self-fold at a bifurcation
• Training Coupled Phase Oscillators as a Neuromorphic Platform using Equilibrium Propagation
• Rigidity of Epithelial Tissues as a Double Optimization Problem
• Intelligent mechanical metamaterials towards learning static and dynamic behaviors
• Self-Contrastive Forward-Forward Algorithm
• Physical Neural Networks with Self-Learning Capabilities
• High-speed coherent photonic random-access memory in long-lasting sound waves
• Physical Computing for Materials Acceleration Platforms
• Measurement-Induced Phase Transitions in Informational Active Matter
• Blending Optimal Control and Biologically Plausible Learning for Noise-Robust Physical Neural Networks
• Physical networks become what they learn
• Kinetic theory of decentralized learning for smart active matter
• A universal approximation theorem for nonlinear resistive networks
• Memory capacity of adaptive flow networks
• Noise-based Local Learning using Stochastic Magnetic Tunnel Junctions
• Unsupervised End-to-End Training with a Self-Defined Target
• Harnessing intuitive local evolution rules for physical learning