Optimizing the energy consumption of spiking neural networks for neuromorphic applications

TL;DR

This paper presents an optimization strategy for training spiking neural networks that reduces energy consumption while maintaining accuracy, leveraging quantization-aware training and sparse computation.

Contribution

It introduces a novel optimization method for spiking neural networks that improves energy efficiency without sacrificing performance.

Findings

Quantization-aware training improves SNN accuracy.

Energy-efficient training maintains accuracy on MNIST-DVS and CIFAR-10.

Sparse computation reduces energy consumption.

Abstract

In the last few years, spiking neural networks have been demonstrated to perform on par with regular convolutional neural networks. Several works have proposed methods to convert a pre-trained CNN to a Spiking CNN without a significant sacrifice of performance. We demonstrate first that quantization-aware training of CNNs leads to better accuracy in SNNs. One of the benefits of converting CNNs to spiking CNNs is to leverage the sparse computation of SNNs and consequently perform equivalent computation at a lower energy consumption. Here we propose an efficient optimization strategy to train spiking networks at lower energy consumption, while maintaining similar accuracy levels. We demonstrate results on the MNIST-DVS and CIFAR-10 datasets.