Research map / Optimization / Optimizers
Differentiable Spike: Rethinking Gradient-Descent for Training Spiking Neural Networks
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 |
|---|---|---|---|
| Dspike adaptive surrogate gradient instead of STBP-tdBN | spiking ResNet · CIFAR10-DVS | top-1 accuracy: 67.8 -> 75.4 | Helped |
| Dspike instead of Diet-SNN | spiking VGG-16 T=5 · ImageNet | top-1 accuracy: 69.00 -> 71.24 | Helped |
| Dspike instead of STBP-tdBN | spiking ResNet-34 T=6 · ImageNet | top-1 accuracy: 63.72 -> 68.19 | Helped |
| Dspike instead of STBP-tdBN at T=4 | spiking ResNet · CIFAR10 | top-1 accuracy: 92.92 -> 93.66 | Helped |
| spiking ResNet-18 T=2 instead of ANN | ResNet-18 · CIFAR10 | energy: 1.03J -> 80.3mJ | Helped |
| use Time Inheritance Training rather than training T=2 from scratch (ablation) | Dspike ResNet-18 · CIFAR10 | accuracy | Helped |
| increase FDG step eps 0.01 -> 0.1 (ablation) | Dspike · CIFAR10 | accuracy: 91.78 -> 94.07 | Helped |
| increase FDG step eps 0.1 -> 0.2 (ablation) | Dspike · CIFAR10/CIFAR100 | accuracy: 94.07 -> 93.84; 74.31 -> 74.57 | Mixed |
| train with FDG eps=0.1 instead of rectangular SG | toy SNN · MNIST | accuracy: 94.87 -> 95.36 | Helped |
| very small FDG step eps=0.001 | toy SNN · MNIST | accuracy: 95.36 -> 91.85 | Hurt |
Closest papers by meaning
- Differentiable hierarchical and surrogate gradient search for spiking neural networks (NeurIPS 2022 · similarity 0.94)
- Take A Shortcut Back: Mitigating the Gradient Vanishing for Training Spiking Neural Networks (NeurIPS 2024 · similarity 0.92)
- Sign Gradient Descent-based Neuronal Dynamics: ANN-to-SNN Conversion Beyond ReLU Network (ICML 2024 · similarity 0.91)
- EnOF-SNN: Training Accurate Spiking Neural Networks via Enhancing the Output Feature (NeurIPS 2024 · similarity 0.90)
- Adaptive Surrogate Gradients for Sequential Reinforcement Learning in Spiking Neural Networks (NeurIPS 2025 · similarity 0.90)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.90)
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.