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

arXiv:2404.15471v1 Announce Type: new
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, …

abstract advances arxiv backpropagation challenges computational cs.lg development energy faster information machine machine learning networks neural networks optical optical neural networks physics.app-ph platforms processing through training type