Training all-mechanical neural networks for task learning through in situ backpropagation
This work addresses the problem of inefficient training in mechanical neural networks, which could advance hardware for machine learning and self-learning materials, though it appears incremental as it adapts existing backpropagation concepts to a new physical domain.
The authors tackled the challenge of training mechanical neural networks (MNNs) by introducing an in situ backpropagation method that computes exact gradients locally, enabling efficient learning. They demonstrated high accuracy in regression and classification tasks and showed retrainability for task-switching and damage resilience.
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.