Research map / Architectures and efficiency / Spiking networks
Intrinsically Stable Spiking Neural Networks: Overcoming the Performance Barrier in the Absence of Batch Normalization
Ruichen Ma, Xiaoyang Zhang, Jian Bai, Guanchao Qiao, Liwei Meng, Ning Ning, Yang Liu, Shaogang Hu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add IS-SNN (topology-aware WS, no BN) instead of no normalization | ResNet-34 · ImageNet (T=4) | accuracy (%): 32.58 -> 66.61 | Improved |
| extend ImageNet training from 128 to 400 epochs | IS-SNN ResNet-34 · ImageNet | accuracy (%): 66.61 -> 68.05 | Improved |
| add IS-SNN to a BN-free SEW-ResNet-152 | ResNet-152 · CIFAR-100 (T=4) | accuracy (%): 17.10 -> 76.98 | Improved |
| add IS-SNN to a BN-free VGG-9 | VGG-9 · CIFAR-10 (T=4) | accuracy (%): 10.00 -> 92.91 | Improved |
| add IS-SNN to a BN-free 7B-Net | 7B-Net · DVS-Gesture (T=16) | accuracy (%): 77.08 -> 96.88 | Improved |
| add IS-SNN to a BN-free VGGSNN | VGGSNN · CIFAR10-DVS (T=8) | accuracy (%): 10.0 -> 77.7 | Improved |
| add IS-SNN to ResNet-19 without BN with absolute gains | ResNet-19 · CIFAR-10 and CIFAR-100 | accuracy (%): +4.49% on CIFAR-10 and +8.58% on CIFAR-100 | Improved |
| use topology-aware gamma instead of gamma fixed to 1 (naive WS) (ablation) | IS-SNN VGG-11 · CIFAR-10 | accuracy (%): 10.00 -> 95.06 | Improved |
| use topology-aware gamma instead of gamma fixed to 1 (naive WS) (ablation) | IS-SNN ResNet-19 · CIFAR-100 | accuracy (%): 70.69 -> 76.02 | Improved |
| use topology-aware gamma instead of gamma fixed to 1 (naive WS) (ablation) | IS-SNN ResNet-19 · CIFAR-10 | accuracy (%): 92.05 -> 94.32 | Improved |
| add IS-SNN to ResNet-19B with five transition blocks (ablation) | ResNet-19B · CIFAR-100 | accuracy (%): 1.00 -> 78.71 | Improved |
| set residual scale alpha=0.5 versus alpha=1.0 (ablation) | IS-SNN ResNet · CIFAR-10/100 | accuracy (%): gap between alpha=0.5 and alpha=1.0 below 0.2% | No clear effect |
Closest papers by meaning
- Training Deep Spiking Neural Networks without Normalization (ICML 2026 · similarity 0.90)
- Robust Stable Spiking Neural Networks (ICML 2024 · similarity 0.89)
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.88)
- Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks (AAAI 2025 · similarity 0.88)
- Are Conventional SNNs Really Efficient? A Perspective from Network Quantization (CVPR 2024 · similarity 0.88)
- Efficient Logit-based Knowledge Distillation of Deep Spiking Neural Networks for Full-Range Timestep Deployment (ICML 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
