Going Deeper in Spiking Neural Networks: VGG and Residual Architectures

TL;DR

This paper introduces a novel method for constructing deep spiking neural networks based on VGG and Residual architectures, achieving high accuracy on complex image recognition tasks while reducing hardware overhead.

Contribution

The paper presents a new algorithmic technique for building deep SNNs that outperform existing methods on CIFAR-10 and ImageNet, applicable to VGG and Residual models.

Findings

Achieved significantly better accuracy than state-of-the-art SNNs.

Demonstrated reduced hardware overhead through analysis of sparse event-driven computations.

Validated effectiveness on complex visual recognition benchmarks.

Abstract

Over the past few years, Spiking Neural Networks (SNNs) have become popular as a possible pathway to enable low-power event-driven neuromorphic hardware. However, their application in machine learning have largely been limited to very shallow neural network architectures for simple problems. In this paper, we propose a novel algorithmic technique for generating an SNN with a deep architecture, and demonstrate its effectiveness on complex visual recognition problems such as CIFAR-10 and ImageNet. Our technique applies to both VGG and Residual network architectures, with significantly better accuracy than the state-of-the-art. Finally, we present analysis of the sparse event-driven computations to demonstrate reduced hardware overhead when operating in the spiking domain.