S2NN: Sub-bit Spiking Neural Networks

Spiking Neural Networks (SNNs) offer an energy-efficient paradigm for machine intelligence, but their continued scaling poses challenges for resource-limited deployment. Despite recent advances in binary SNNs, the storage and computational demands remain substantial for large-scale networks. To further explore the compression and acceleration potential of SNNs, we propose Sub-bit Spiking Neural Networks (S2NNs) that represent weights with less than one bit. Specifically, we first establish an S2NN baseline by leveraging the clustering patterns of kernels in well-trained binary SNNs. This baseline is highly efficient but suffers from textitoutlier-induced codeword selection bias during training. To mitigate this issue, we propose an textitoutlier-aware sub-bit weight quantization (OS-Quant) method, which optimizes codeword selection by identifying and adaptively scaling outliers. Furthermore, we propose a textitmembrane potential-based feature distillation (MPFD) method, improving the performance of highly compressed S2NN via more precise guidance from a teacher model. Extensive results on vision tasks reveal that S2NN outperforms existing quantized SNNs in both performance and efficiency, making it promising for edge computing applications.