Research map / Architectures and efficiency / Spiking networks
Breaking Gradient Temporal Collinearity for Robust Spiking Neural Networks
Desong Zhang, Jia Hu, Geyong Min
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 |
|---|---|---|---|
| add orthogonal kernels at inference (ablation) | STOD · CIFAR-10 PGD | accuracy %: 32.97 -> 36.72 | Improved |
| add orthogonal kernels at inference (ablation) | STOD · CIFAR-10 FGSM | accuracy %: 55.80 -> 59.16 | Improved |
| add orthogonal kernels at inference (ablation) | STOD · CIFAR-100 FGSM | accuracy %: 26.26 -> 32.15 | Improved |
| add orthogonal kernels at inference (ablation) | STOD · ImageNet PGD | accuracy %: 6.44 -> 8.01 | Improved |
| add orthogonal kernels at inference (ablation) | STOD · CIFAR-10 clean | accuracy %: 91.43 -> 90.87 | Hurt |
| replace direct encoding with STOD temporal decorrelated input encoding | SNN · CIFAR-10 clean | accuracy %: 93.75 -> 91.43 | Hurt |
| replace direct encoding with STOD | SNN · CIFAR-10 PGD | accuracy %: 0.03 -> 32.97 | Improved |
| replace direct encoding with STOD | SNN · CIFAR-100 PGD | accuracy %: 0.19 -> 13.13 | Improved |
| replace direct encoding with STOD | SNN · DVS-CIFAR10 FGSM | accuracy %: 17.20 -> 66.10 | Improved |
| replace direct encoding with STOD under RGA attack | SNN · CIFAR-10 white-box RGA PGD | accuracy %: 1.01 -> 49.40 | Improved |
| add STOD (no OK) on top of prior SOTA SR on DVS-CIFAR10 | SR · DVS-CIFAR10 PGD | accuracy %: 61.20 -> 60.40 | No clear effect |
| increase PGD steps from 7 to 100 | STOD · CIFAR-10 PGD | accuracy %: 36.72 -> 30.93 | Hurt |
Closest papers by meaning
- Training High Performance Spiking Neural Network by Temporal Model Calibration (ICML 2025 · similarity 0.89)
- Rethinking Spiking Neural Networks from an Ensemble Learning Perspective (ICLR 2025 · similarity 0.89)
- FSTA-SNN:Frequency-Based Spatial-Temporal Attention Module for Spiking Neural Networks (AAAI 2025 · similarity 0.89)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.88)
- Unveiling the Spatial-temporal Effective Receptive Fields of Spiking Neural Networks (NeurIPS 2025 · similarity 0.88)
- STAA-SNN: Spatial-Temporal Attention Aggregator for Spiking Neural Networks (CVPR 2025 · similarity 0.88)
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.
