AI Paper Atlas

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

ICLR 2026 · Architectures and efficiency › Spiking networks · approach: Spiking transformers · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add DSD (deflate dominant singular component of gradients)vanilla SNN (BPTT, direct encoding) · CIFAR-100 vanilla training, FGSMaccuracy: 4.55 -> 23.81Improved
add DSDvanilla SNN · CIFAR-10 vanilla training, PGDaccuracy: 0.03 -> 31.44Improved
add DSD in adversarial trainingadversarially trained SNN · TinyImageNet AT, FGSMaccuracy: 8.19 -> 30.87Improved
add DSDvanilla SNN · CIFAR-10 vanilla training, clean accuracyaccuracy: 93.75 -> 90.21Hurt
add DSDvanilla SNN · CIFAR-100 vanilla training, PGDaccuracy: 2.04 -> 8.09Improved
add DSDvanilla SNN · CIFAR-10 hetero-training c/+p_0 with b=1, FGSMaccuracy: 0.34 -> 33.70Improved
add DSDvanilla SNN · CIFAR-10 Hessian spectral radius, clean inferencelambda_max(H): 261.94 -> 209.90Improved
add DSDvanilla SNN · CIFAR-10 APGD-CE vanilla trainingaccuracy: 0.61 -> 22.19Improved
add DSDSNN with adversarial training · DVS-Gesture, FGSMaccuracy: 39.24 -> 90.28Improved
replace BPTT with SLTT (ablation)SNN under hetero-training · CIFAR-10 hetero-training collapseaccuracyImproved
replace direct encoding with rate encoding (ablation)SNN under hetero-training · CIFAR-10 hetero-training collapseaccuracyImproved
increase number of time steps (ablation)direct-encoding VGG-11 SNN · CIFAR-10 hetero-trainingaccuracyHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy