ApproxDARTS: Differentiable Neural Architecture Search with Approximate Multipliers

arXiv:2404.08002v1 Announce Type: new
Abstract: Integrating the principles of approximate computing into the design of hardware-aware deep neural networks (DNN) has led to DNNs implementations showing good output quality and highly optimized hardware parameters such as low latency or inference energy. In this work, we present ApproxDARTS, a neural architecture search (NAS) method enabling the popular differentiable neural architecture search method called DARTS to exploit approximate multipliers and thus reduce the power consumption of generated neural networks. We showed on …

abstract architecture arxiv computing cs.lg design differentiable dnn energy good hardware inference latency low low latency nas networks neural architecture search neural networks parameters quality search type work