HW/SW Codesign for Robust and Efficient Binarized SNNs by Capacitor Minimization

TL;DR

This paper introduces CapMin, a hardware/software co-design approach that significantly reduces capacitor size in analog IF-SNNs, improving energy efficiency and robustness against variations.

Contribution

CapMin is a novel method that minimizes capacitor size by reducing spike times based on frequency analysis, enhancing analog SNN efficiency.

Findings

Over 14x reduction in capacitor size compared to state-of-the-art

CapMin-V increases variation tolerance with minimal capacitor size increase

Improves energy efficiency and robustness of analog IF-SNNs

Abstract

Using accelerators based on analog computing is an efficient way to process the immensely large workloads in Neural Networks (NNs). One example of an analog computing scheme for NNs is Integrate-and-Fire (IF) Spiking Neural Networks (SNNs). However, to achieve high inference accuracy in IF-SNNs, the analog hardware needs to represent current-based multiply-accumulate (MAC) levels as spike times, for which a large membrane capacitor needs to be charged for a certain amount of time. A large capacitor results in high energy use, considerable area cost, and long latency, constituting one of the major bottlenecks in analog IF-SNN implementations. In this work, we propose a HW/SW Codesign method, called CapMin, for capacitor size minimization in analog computing IF-SNNs. CapMin minimizes the capacitor size by reducing the number of spike times needed for accurate operation of the HW, based on the absolute frequency of MAC level occurrences in the SW. To increase the operation of IF-SNNs to current variation, we propose the method CapMin-V, which trades capacitor size for protection based on the reduced capacitor size found in CapMin. In our experiments, CapMin achieves more than a 14$\times$ reduction in capacitor size over the state of the art, while CapMin-V achieves increased variation tolerance in the IF-SNN operation, requiring only a small increase in capacitor size.