Research map / Architectures and efficiency / Spiking networks
Exploring Loss Functions for Time-based Training Strategy in Spiking Neural Networks
Yaoyu Zhu, Wei Fang, Xiaodong Xie, Tiejun Huang, Zhaofei Yu
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 counting (MSE) loss with enhanced counting loss, weight normalization (ablation) | time-based SNN (EDBP baseline) · CIFAR-100 | top-1 accuracy: 63.98% -> 68.95% | Improved |
| Replace counting (MSE) loss with enhanced counting loss, threshold normalization (ablation) | time-based SNN · CIFAR-100 | top-1 accuracy: 66.35% -> 70.50% | Improved |
| Replace weight normalization with threshold normalization, counting loss (ablation) | time-based SNN · CIFAR-100 | top-1 accuracy: 63.98% -> 66.35% | Improved |
| Replace weight normalization with threshold normalization, enhanced loss (ablation) | time-based SNN · CIFAR-100 | top-1 accuracy: 68.95% -> 70.50% | Improved |
| Use enhanced counting loss (ablation) | time-based SNN · CIFAR-100 | gradient scale (L1 norm) | Improved |
| Use threshold normalization (ablation) | time-based SNN · CIFAR-100 | gradient scale | Improved |
| Use enhanced loss with threshold normalization (ablation) | time-based SNN · CIFAR-100 | ratio of positive gradients | Improved |
| Use proposed method instead of EDBP with VGG11 | EDBP · CIFAR10 | top-1 accuracy: 92.10% -> 93.54% | Improved |
| Use proposed method instead of EDBP with VGG11 | EDBP · CIFAR100 | top-1 accuracy: 63.97% -> 70.50% | Improved |
| Use proposed method on saturated dataset | TSSL-BP · MNIST | top-1 accuracy: 99.53% -> 99.50% | No clear effect |
| Use proposed method instead of EDBP | EDBP · Fashion-MNIST | top-1 accuracy: 93.28% -> 94.03% | Improved |
| Use proposed method instead of EDBP | EDBP · N-MNIST | top-1 accuracy: 99.39% -> 99.39% | No clear effect |
Closest papers by meaning
- Efficiently Training Time-to-First-Spike Spiking Neural Networks from Scratch (ICML 2026 · similarity 0.92)
- Temporal Weighted Encoding: Towards Maximal-Capacity Spike Coding for ANN–SNN Conversion (ICML 2026 · similarity 0.91)
- Advancing Training Efficiency of Deep Spiking Neural Networks through Rate-based Backpropagation (NeurIPS 2024 · similarity 0.91)
- Temporal Efficient Training of Spiking Neural Network via Gradient Re-weighting (ICLR 2022 · similarity 0.91)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.90)
- TS-SNN: Temporal Shift Module for Spiking Neural Networks (ICML 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.
