Training all-mechanical neural networks for task learning through in situ backpropagation

TL;DR

This paper introduces a novel in situ backpropagation method for mechanical neural networks, enabling efficient training and demonstrating their potential for autonomous learning and adaptability in hardware systems.

Contribution

The paper develops the first mechanical analogue of in situ backpropagation, allowing local gradient computation and training of mechanical neural networks for various learning tasks.

Findings

Successfully trained MNNs for regression and classification tasks.

Demonstrated task-switching and damage resilience in MNNs.

Validated the approach through experimental and numerical results.

Abstract

Recent advances unveiled physical neural networks as promising machine learning platforms, offering faster and more energy-efficient information processing. Compared with extensively-studied optical neural networks, the development of mechanical neural networks (MNNs) remains nascent and faces significant challenges, including heavy computational demands and learning with approximate gradients. Here, we introduce the mechanical analogue of in situ backpropagation to enable highly efficient training of MNNs. We demonstrate that the exact gradient can be obtained locally in MNNs, enabling learning through their immediate vicinity. With the gradient information, we showcase the successful training of MNNs for behavior learning and machine learning tasks, achieving high accuracy in regression and classification. Furthermore, we present the retrainability of MNNs involving task-switching and damage, demonstrating the resilience. Our findings, which integrate the theory for training MNNs and experimental and numerical validations, pave the way for mechanical machine learning hardware and autonomous self-learning material systems.