Quantization of Spiking Neural Networks Beyond Accuracy

TL;DR

This paper highlights the importance of preserving firing behavior in quantized spiking neural networks and introduces Earth Mover's Distance as a diagnostic metric for this purpose.

Contribution

It proposes using EMD to evaluate firing distribution divergence in quantized SNNs and demonstrates its effectiveness across different quantization methods.

Findings

Uniform quantization causes distributional drift despite accuracy retention.

LQ-Net style learned quantization maintains firing behavior similar to full-precision.

EMD is a useful tool for behavior preservation assessment in SNN quantization.

Abstract

Quantization is a natural complement to the sparse, event-driven computation of Spiking Neural Networks, reducing memory bandwidth and arithmetic cost for deployment on resource-constrained hardware. However, existing SNN quantization evaluation focuses almost exclusively on accuracy, overlooking whether a quantized network preserves the firing behavior of its full-precision counterpart. We demonstrate that quantization method, clipping range, and bit-width can produce substantially different firing distributions at equivalent accuracy, differences invisible to standard metrics but relevant to deployment, where firing activity governs effective sparsity, state storage, and event-processing load. To capture this gap, we propose Earth Mover's Distance as a diagnostic metric for firing distribution divergence, and apply it systematically across weight and membrane quantization on SEW-ResNet architectures trained on CIFAR-10 and CIFAR-100. We find that uniform quantization induces distributional drift even when accuracy is preserved, while LQ-Net style learned quantization maintains firing behavior close to the full-precision baseline. Our results suggest that behavior preservation should be treated as an evaluation criterion alongside accuracy, and that EMD provides a principled tool for assessing it.