Research map / Architectures and efficiency / Spiking networks
Temporal Efficient Training of Spiking Neural Network via Gradient Re-weighting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 standard direct training loss with per-timestep TET loss | ResNet-19 SNN with surrogate gradients · CIFAR100, T=2 | accuracy: 69.41$\pm$0.08 -> 72.37$\pm$0.21 | Helped |
| Replace standard direct training loss with per-timestep TET loss | ResNet-19 SNN with surrogate gradients · CIFAR100, T=4 | accuracy: 70.86$\pm$0.22 -> 74.11$\pm$0.18 | Helped |
| Replace standard direct training loss with per-timestep TET loss | ResNet-19 SNN with surrogate gradients · CIFAR100, T=6 | accuracy: 71.12$\pm$0.57 -> 74.65$\pm$0.12 | Helped |
| Use TET rather than standard direct training with sigmoid activation (ablation) | CNN-5 · CIFAR10, k=1 | accuracy: 88.00$\pm$0.15 -> 87.63$\pm$0.38 | No clear effect |
| Use TET rather than standard direct training with sigmoid activation (ablation) | CNN-5 · CIFAR10, k=10 | accuracy: 88.83$\pm$0.32 -> 89.31$\pm$0.15 | No clear effect |
| Use TET rather than standard direct training with sigmoid activation (ablation) | CNN-5 · CIFAR10, k=20 | accuracy: 88.50$\pm$0.32 -> 88.64$\pm$0.28 | No clear effect |
| Add the MSE regularization term to TET (ablation) | ResNet-19 · CIFAR100 | accuracy | Helped |
| Add the MSE regularization term to TET (ablation) | VGGSNN · DVS-CIFAR10 | accuracy | Helped |
| Use a smaller MSE regularization weight rather than a large one (ablation) | VGGSNN trained with TET · DVS-CIFAR10 | accuracy | Helped |
| Initialize with a short simulation length, then increase length and fine-tune (TIT) (ablation) | TET-trained SNN · CIFAR100 | accuracy | Helped |
| Initialize with a short simulation length, then increase length and fine-tune (TIT) (ablation) | TET-trained SNN · training time | training time: 1800$ts$ -> 900$ts$ | Helped |
| Train with TET rather than SDT before expanding simulation length (ablation) | ResNet-19 · CIFAR100, simulation length expanded to 64 | accuracy variance: 11.13 -> 1.80 | Helped |
Closest papers by meaning
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.93)
- NDOT: Neuronal Dynamics-based Online Training for Spiking Neural Networks (ICML 2024 · similarity 0.92)
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.91)
- Advancing Training Efficiency of Deep Spiking Neural Networks through Rate-based Backpropagation (NeurIPS 2024 · similarity 0.91)
- Take A Shortcut Back: Mitigating the Gradient Vanishing for Training Spiking Neural Networks (NeurIPS 2024 · similarity 0.91)
- EnOF-SNN: Training Accurate Spiking Neural Networks via Enhancing the Output Feature (NeurIPS 2024 · similarity 0.91)
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.