Layer Ensemble Averaging for Improving Memristor-Based Artificial Neural Network Performance

arXiv:2404.15621v1 Announce Type: cross
Abstract: Artificial neural networks have advanced due to scaling dimensions, but conventional computing faces inefficiency due to the von Neumann bottleneck. In-memory computation architectures, like memristors, offer promise but face challenges due to hardware non-idealities. This work proposes and experimentally demonstrates layer ensemble averaging, a technique to map pre-trained neural network solutions from software to defective hardware crossbars of emerging memory devices and reliably attain near-software performance on inference. The approach is investigated using a custom …

abstract advanced architectures artificial artificial neural networks arxiv challenges computation computing cs.ar cs.et cs.lg dimensions eess.iv ensemble face hardware improving in-memory layer memory memristor network networks neural network neural networks performance scaling type work