Real Spike: Learning Real-valued Spikes for Spiking Neural Networks

TL;DR

This paper introduces Real Spike, a novel training-inference decoupling method for SNNs that uses unshared convolution kernels and binary spikes during inference, leading to improved performance and hardware compatibility.

Contribution

It proposes a re-parameterization technique enabling unshared kernels and binary spikes in inference while maintaining shared kernels during training, enhancing SNN efficiency.

Findings

Real Spike improves SNN performance across multiple datasets.

The method outperforms state-of-the-art models on static and neuromorphic datasets.

Theoretical proof confirms the superiority of Real Spike-based SNNs.

Abstract

Brain-inspired spiking neural networks (SNNs) have recently drawn more and more attention due to their event-driven and energy-efficient characteristics. The integration of storage and computation paradigm on neuromorphic hardwares makes SNNs much different from Deep Neural Networks (DNNs). In this paper, we argue that SNNs may not benefit from the weight-sharing mechanism, which can effectively reduce parameters and improve inference efficiency in DNNs, in some hardwares, and assume that an SNN with unshared convolution kernels could perform better. Motivated by this assumption, a training-inference decoupling method for SNNs named as Real Spike is proposed, which not only enjoys both unshared convolution kernels and binary spikes in inference-time but also maintains both shared convolution kernels and Real-valued Spikes during training. This decoupling mechanism of SNN is realized by a re-parameterization technique. Furthermore, based on the training-inference-decoupled idea, a series of different forms for implementing Real Spike on different levels are presented, which also enjoy shared convolutions in the inference and are friendly to both neuromorphic and non-neuromorphic hardware platforms. A theoretical proof is given to clarify that the Real Spike-based SNN network is superior to its vanilla counterpart. Experimental results show that all different Real Spike versions can consistently improve the SNN performance. Moreover, the proposed method outperforms the state-of-the-art models on both non-spiking static and neuromorphic datasets.