ReverB-SNN: Reversing Bit of the Weight and Activation for Spiking Neural Networks

TL;DR

ReverB-SNN introduces a novel approach to improve spiking neural networks by reversing bits of weights and activations, using real-valued activations and trainable weights to enhance accuracy while maintaining efficiency.

Contribution

The paper proposes ReverB-SNN, which reverses bits of weights and activations, employs real-valued activations, and introduces trainable weights for improved accuracy in SNNs.

Findings

Outperforms state-of-the-art SNN methods across multiple datasets.

Maintains energy efficiency with real-valued activations and binary weights.

Enhances information capacity of activations without sacrificing efficiency.

Abstract

The Spiking Neural Network (SNN), a biologically inspired neural network infrastructure, has garnered significant attention recently. SNNs utilize binary spike activations for efficient information transmission, replacing multiplications with additions, thereby enhancing energy efficiency. However, binary spike activation maps often fail to capture sufficient data information, resulting in reduced accuracy. To address this challenge, we advocate reversing the bit of the weight and activation for SNNs, called \textbf{ReverB-SNN}, inspired by recent findings that highlight greater accuracy degradation from quantizing activations compared to weights. Specifically, our method employs real-valued spike activations alongside binary weights in SNNs. This preserves the event-driven and multiplication-free advantages of standard SNNs while enhancing the information capacity of activations. Additionally, we introduce a trainable factor within binary weights to adaptively learn suitable weight amplitudes during training, thereby increasing network capacity. To maintain efficiency akin to vanilla \textbf{ReverB-SNN}, our trainable binary weight SNNs are converted back to standard form using a re-parameterization technique during inference. Extensive experiments across various network architectures and datasets, both static and dynamic, demonstrate that our approach consistently outperforms state-of-the-art methods.