Examining the Robustness of Spiking Neural Networks on Non-ideal Memristive Crossbars

TL;DR

This paper investigates how non-idealities in memristive crossbars affect the performance of Spiking Neural Networks, revealing error accumulation issues and the benefits of fewer time-steps for training.

Contribution

It provides a comprehensive analysis of SNN robustness on non-ideal memristive crossbars, highlighting error accumulation and training strategies for improved accuracy.

Findings

Error accumulation during crossbar computations causes significant performance drops.

Training SNNs with fewer time-steps improves accuracy on memristive crossbars.

Repetitive crossbar operations across multiple time-steps exacerbate errors.

Abstract

Spiking Neural Networks (SNNs) have recently emerged as the low-power alternative to Artificial Neural Networks (ANNs) owing to their asynchronous, sparse, and binary information processing. To improve the energy-efficiency and throughput, SNNs can be implemented on memristive crossbars where Multiply-and-Accumulate (MAC) operations are realized in the analog domain using emerging Non-Volatile-Memory (NVM) devices. Despite the compatibility of SNNs with memristive crossbars, there is little attention to study on the effect of intrinsic crossbar non-idealities and stochasticity on the performance of SNNs. In this paper, we conduct a comprehensive analysis of the robustness of SNNs on non-ideal crossbars. We examine SNNs trained via learning algorithms such as, surrogate gradient and ANN-SNN conversion. Our results show that repetitive crossbar computations across multiple time-steps induce error accumulation, resulting in a huge performance drop during SNN inference. We further show that SNNs trained with a smaller number of time-steps achieve better accuracy when deployed on memristive crossbars.