Research map / Architectures and efficiency / Spiking networks
SSF: Accelerating Training of Spiking Neural Networks with Stabilized Spiking Flow
Jingtao Wang, Zengjie Song, Yuxi Wang, Jun Xiao, Yuran Yang, Shuqi Mei, Zhaoxiang Zhang
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 |
|---|---|---|---|
| replace backprop through membrane accumulation with stabilized spiking flow mapping | TET, RecDis, SEW, tdBN · CIFAR-10/CIFAR-100/TinyImageNet (T=20) | backward time speed-up: 2.01x to 2.95x speed-up with accuracy on par | Improved |
| apply SSF to TET on neuromorphic data | TET · DVS-CIFAR10 (T=20) | accuracy: 78.7% -> 74.1% | Hurt |
| swap TET loss for cross-entropy loss when using SSF (ablation) | TET + SSF · DVS-CIFAR10 (T=20) | accuracy: 74.1% -> 78.0% | Improved |
| use learned average spiking electricity R instead of fixed R=1 (ablation) | SSF · CIFAR-10 | accuracy: 10.00% -> 95.19% | Improved |
| use fine-grained R (per element) instead of coarser R (per batch or per batch and channel) (ablation) | SSF · CIFAR-10 | accuracy: 60.19% -> 95.19% | Improved |
| use a single scalar R instead of per-element R (ablation) | SSF · CIFAR-10 | accuracy: 95.19% -> 94.41% | Hurt |
| SSF instead of DSR without firing mechanism modification, fewer time steps | IF SNN · CIFAR-10 (T=5) | accuracy: 84.0% -> 95.0% | Improved |
| increase number of GPUs from 1 to 8 (ablation) | SSF speed-up · TET on CIFAR-10 | speed-up ratio: 3.04 -> 2.96 | No clear effect |
| increase time steps up to 80 with SSF | SSF · TET training time | speed-up ratio: about 10x at T=80 | Improved |
Closest papers by meaning
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.91)
- Take A Shortcut Back: Mitigating the Gradient Vanishing for Training Spiking Neural Networks (NeurIPS 2024 · similarity 0.90)
- EnOF-SNN: Training Accurate Spiking Neural Networks via Enhancing the Output Feature (NeurIPS 2024 · similarity 0.90)
- Efficient Parallel Training Methods for Spiking Neural Networks with Constant Time Complexity (ICML 2025 · similarity 0.90)
- Rethinking the Membrane Dynamics and Optimization Objectives of Spiking Neural Networks (NeurIPS 2024 · similarity 0.90)
- Adaptive Surrogate Gradients for Sequential Reinforcement Learning in Spiking Neural Networks (NeurIPS 2025 · similarity 0.89)
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.
