Research map / Architectures and efficiency / Spiking networks
Differentiable hierarchical and surrogate gradient search for 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 |
|---|---|---|---|
| add differentiable surrogate gradient search (DGS) to SpikeDHS-CLA (n3c5 vs n3) (ablation) | SpikeDHS-CLA · CIFAR10 | accuracy (%): 95.35 -> 95.50 | Helped |
| add DGS on first stem layer (n4) (ablation) | SpikeDHS-CLA · CIFAR10 | accuracy (%): 94.34 -> 94.68 | Helped |
| add DGS on first stem layer (n3) (ablation) | SpikeDHS-CLA · CIFAR100 | accuracy (%): 76.15 -> 76.25 | No clear effect |
| add DGS to SpikeDHS-CLA-large (ablation) | SpikeDHS-CLA-large · ImageNet | accuracy (%): 67.96 -> 68.64 | Helped |
| add DGS to SpikeDHS-Stereo (ablation) | SpikeDHS-Stereo · MVSEC split1 | mean depth error (cm): 16.5 -> 15.9 | Helped |
| replace binary neurons (tau=0) with LIF temporal dynamics (ablation) | SpikeDHS-Stereo · MVSEC split3 | mean depth error (cm): 20.3 -> 19.4 | Helped |
| mix operations at membrane potential input (MM) instead of spike output (ablation) | SpikeDHS-Stereo · MVSEC split1 | mean depth error (cm): 16.5 -> 15.7 | Helped |
| add DGS to Triangle SG (ablation) | SpikeDHS-Stereo · MVSEC split1 | 1PA (%): 90.9 -> 91.3 | Helped |
| add DGS to Arctan SG (ablation) | SpikeDHS-Stereo · MVSEC split1 | 1PA (%): 89.3 -> 90.1 | Helped |
| add DGS to Superspike SG (ablation) | SpikeDHS-Stereo · MVSEC split1 | 1PA (%): 89.6 -> 89.7 | No clear effect |
| use searched SNN architecture instead of hand-designed U-net (StereoSpike) | event-based stereo SNN · MVSEC split1 | mean depth error (cm): 18.5 -> 16.5 | Helped |
Closest papers by meaning
- Differentiable Spike: Rethinking Gradient-Descent for Training Spiking Neural Networks (NeurIPS 2021 · similarity 0.94)
- Temporal Efficient Training of Spiking Neural Network via Gradient Re-weighting (ICLR 2022 · similarity 0.92)
- A$^2$SG: Adaptive and Asymmetric Surrogate Gradients for Training Deep Spiking Neural Networks (ICML 2026 · 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 0.90)
- S$^2$NN: Sub-bit 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.