Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2604.25688 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908998989512704 |
|---|---|
| author | Bai, Dewei Peng, Hongxiang Mei, Jiajun Ren, Yang Qu, Hong Xia, Dawen Yi, Zhang |
| author_facet | Bai, Dewei Peng, Hongxiang Mei, Jiajun Ren, Yang Qu, Hong Xia, Dawen Yi, Zhang |
| contents | Binary spike coding enables sparse and event-driven computation in spiking neural networks (SNNs), yet its 1-bit-per-timestep representation fundamentally limits information throughput. This bottleneck becomes increasingly restrictive in deep architectures under short simulation horizons. We propose the Quantized Burst-LIF (QB-LIF) neuron, which reformulates burst spiking as a saturated uniform quantization of membrane potentials with a learnable scale. Instead of relying on predefined multi-threshold structures, QB-LIF treats the quantization scale as a trainable parameter, allowing each layer to autonomously adapt its spiking resolution to the underlying membrane-potential statistics. To preserve hardware efficiency, we introduce an absorbable scale strategy that folds the learned quantized scale into synaptic weights during inference, maintaining a strict accumulate-only (AC) execution paradigm. To enable stable optimization in the discrete multi-level space, we further design ReLSG-ET, a rectified-linear surrogate gradient with exponential tails that sustains gradient flow across burst intervals. Extensive experiments on static (CIFAR-10/100, ImageNet) and event-driven (CIFAR10-DVS, DVS128-Gesture) benchmarks demonstrate that QB-LIF consistently outperforms binary and fixed-burst SNNs, achieving higher accuracy under ultra-low latency while preserving neuromorphic compatibility. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_25688 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | QB-LIF: Learnable-Scale Quantized Burst Neurons for Efficient SNNs Bai, Dewei Peng, Hongxiang Mei, Jiajun Ren, Yang Qu, Hong Xia, Dawen Yi, Zhang Computer Vision and Pattern Recognition Binary spike coding enables sparse and event-driven computation in spiking neural networks (SNNs), yet its 1-bit-per-timestep representation fundamentally limits information throughput. This bottleneck becomes increasingly restrictive in deep architectures under short simulation horizons. We propose the Quantized Burst-LIF (QB-LIF) neuron, which reformulates burst spiking as a saturated uniform quantization of membrane potentials with a learnable scale. Instead of relying on predefined multi-threshold structures, QB-LIF treats the quantization scale as a trainable parameter, allowing each layer to autonomously adapt its spiking resolution to the underlying membrane-potential statistics. To preserve hardware efficiency, we introduce an absorbable scale strategy that folds the learned quantized scale into synaptic weights during inference, maintaining a strict accumulate-only (AC) execution paradigm. To enable stable optimization in the discrete multi-level space, we further design ReLSG-ET, a rectified-linear surrogate gradient with exponential tails that sustains gradient flow across burst intervals. Extensive experiments on static (CIFAR-10/100, ImageNet) and event-driven (CIFAR10-DVS, DVS128-Gesture) benchmarks demonstrate that QB-LIF consistently outperforms binary and fixed-burst SNNs, achieving higher accuracy under ultra-low latency while preserving neuromorphic compatibility. |
| title | QB-LIF: Learnable-Scale Quantized Burst Neurons for Efficient SNNs |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.25688 |