Research map / Architectures and efficiency / Spiking networks
Robustify Spiking Neural Networks via Dominant Singular Deflation under Heterogeneous Training Vulnerability
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 DSD (deflate dominant singular component of gradients) | vanilla SNN (BPTT, direct encoding) · CIFAR-100 vanilla training, FGSM | accuracy: 4.55 -> 23.81 | Improved |
| add DSD | vanilla SNN · CIFAR-10 vanilla training, PGD | accuracy: 0.03 -> 31.44 | Improved |
| add DSD in adversarial training | adversarially trained SNN · TinyImageNet AT, FGSM | accuracy: 8.19 -> 30.87 | Improved |
| add DSD | vanilla SNN · CIFAR-10 vanilla training, clean accuracy | accuracy: 93.75 -> 90.21 | Hurt |
| add DSD | vanilla SNN · CIFAR-100 vanilla training, PGD | accuracy: 2.04 -> 8.09 | Improved |
| add DSD | vanilla SNN · CIFAR-10 hetero-training c/+p_0 with b=1, FGSM | accuracy: 0.34 -> 33.70 | Improved |
| add DSD | vanilla SNN · CIFAR-10 Hessian spectral radius, clean inference | lambda_max(H): 261.94 -> 209.90 | Improved |
| add DSD | vanilla SNN · CIFAR-10 APGD-CE vanilla training | accuracy: 0.61 -> 22.19 | Improved |
| add DSD | SNN with adversarial training · DVS-Gesture, FGSM | accuracy: 39.24 -> 90.28 | Improved |
| replace BPTT with SLTT (ablation) | SNN under hetero-training · CIFAR-10 hetero-training collapse | accuracy | Improved |
| replace direct encoding with rate encoding (ablation) | SNN under hetero-training · CIFAR-10 hetero-training collapse | accuracy | Improved |
| increase number of time steps (ablation) | direct-encoding VGG-11 SNN · CIFAR-10 hetero-training | accuracy | Hurt |
Closest papers by meaning
- Enhanced Self-Distillation Framework for Efficient Spiking Neural Network Training (NeurIPS 2025 · similarity 0.89)
- Synergy Between the Strong and the Weak: Spiking Neural Networks are Inherently Self-Distillers (NeurIPS 2025 · similarity 0.88)
- Robust Stable Spiking Neural Networks (ICML 2024 · similarity 0.88)
- Temporal Flexibility in Spiking Neural Networks: Towards Generalization Across Time Steps and Deployment Friendliness (ICLR 2025 · similarity 0.88)
- Enhancing Adversarial Robustness in SNNs with Sparse Gradients (ICML 2024 · similarity 0.87)
- Take A Shortcut Back: Mitigating the Gradient Vanishing for Training Spiking Neural Networks (NeurIPS 2024 · similarity 0.87)
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.
