Research map / Adversarial robustness and security / Certified and empirical robustness
A Unified Total Variation Framework for Membrane Potential Perturbation Dynamic
Zhao-Rong Lai, Xiwen Yuan, Ziliang Chen, Liangda Fang, Yongsen Zheng
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 |
|---|---|---|---|
| replace squared TV-l2 MPPD regularizer with TV-l1 | MPPD SNN, VGG11 AT · CIFAR-10 PGD7 | accuracy: 34.510 -> 42.840 | Improved |
| replace TV-l2 with TV-l1 | MPPD SNN, VGG11 AT · CIFAR-10 AutoAttack | accuracy: 19.75 -> 23.040 | Improved |
| replace TV-l2 with TV-l1 | MPPD SNN, VGG11 Gaussian · CIFAR-100 PGD7 | accuracy: 0.690 -> 1.390 | Improved |
| replace TV-l2 with TV-l1 | MPPD SNN, WRN16 AT · CIFAR-10 AutoAttack | accuracy: 21.340 -> 22.920 | Improved |
| replace TV-l2 with TV-l1 | MPPD SNN, VGG11 AT · Tiny ImageNet AutoAttack | accuracy: 3.890 -> 4.080 | Improved |
| add MPPD-TV-l1 regularizer versus none | SNN VGG11 AT · CIFAR-10 PGD7 | accuracy: 35.520 -> 42.840 | Improved |
| add MPPD-TV-l1 regularizer on top of AT+Reg | SNN VGG11 AT+Reg · CIFAR-10 AutoAttack | accuracy: 14.06 -> 23.530 | Improved |
| add MPPD-TV-l1 regularizer on top of AT+Reg | SNN VGG11 AT+Reg · CIFAR-100 clean | accuracy: 62.190 -> 62.710 | No clear effect |
| add TV-l2 MPPD regularizer to standard SNN | SNN VGG11 AT · CIFAR-100 PGD7 | accuracy: 19.950 -> 16.440 | Hurt |
| add TV-l1 regularizer alpha=1.0 versus alpha=0 (ablation) | MPPD-TV-l1 VGG5 AT · CIFAR-10 AutoAttack | accuracy: 18.870 -> 22.270 | Improved |
| increase alpha from 0.0 to 2.5 (ablation) | MPPD-TV-l1 VGG5 AT · CIFAR-10 PGD7 | accuracy: 30.990 -> 34.690 | Improved |
| increase alpha from 3.0 to 3.5 (ablation) | MPPD-TV-l1 VGG5 AT · CIFAR-10 AutoAttack | accuracy: 25.020 -> 13.590 | Hurt |
Closest papers by meaning
- Robust Stable Spiking Neural Networks (ICML 2024 · similarity 0.89)
- MPD-SGR: Robust Spiking Neural Networks with Membrane Potential Distribution-Driven Surrogate Gradient Regularization (AAAI 2026 · similarity 0.89)
- Optimization Method for Surrogate Function in Spiking Neural Networks Based on Membrane Potential Distribution (AAAI 2026 · similarity 0.87)
- Do Not Let Spikes Flip: Margin-Resculpted Learning for Robust Spiking Neural Networks (NeurIPS 2026 · similarity 0.87)
- Rethinking the Membrane Dynamics and Optimization Objectives of Spiking Neural Networks (NeurIPS 2024 · similarity 0.85)
- Rethinking Spiking Neural Networks from an Ensemble Learning Perspective (ICLR 2025 · similarity 0.84)
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.
