AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Robust Spiking Neural Networks by Temporal Mutual Information

Mengting Xu, Shi Gu, Peng Lin, De Ma, Huajin Tang, Qian Zheng, Gang Pan

CVPR 2026 · Architectures and efficiency › Spiking networks · approach: Training deep spiking networks

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 TMI regularizer to STBP trainingSTBP VGGSNN · CIFAR-100, FGSMaccuracy (%): 13.51 -> 15.20Improved
add TMI regularizer to TET trainingTET VGGSNN · CIFAR-100, PGDaccuracy (%): 2.25 -> 4.63Improved
add TMI regularizer to SNN-RAT adversarial trainingSNN-RAT VGG11 · CIFAR-100, FGSMaccuracy (%): 25.86 -> 27.32Improved
add TMI regularizerSTBP VGGSNN · CIFAR-100, natural accuracyaccuracy (%): 68.11 -> 68.12No clear effect
add TMI regularizerTET VGGSNN · CIFAR-100, natural accuracyaccuracy (%): 72.66 -> 72.32Hurt
add output-entropy regularizer (H) instead of TMISTBP VGGSNN · CIFAR-100, FGSMaccuracy (%): 13.51 -> 12.14Hurt
estimate MI with histogram (AIMIE) instead of KDESTBP VGGSNN · CIFAR-100, FGSMaccuracy (%): 13.51 -> 12.54Hurt
use max-pooling over channels instead of averaging in TMI (ablation)TMI regularizer · CIFAR-10 VGGSNN, FGSMaccuracy (%): 12.23 -> 25.80Improved
compute MI on the last spike layer rather than a shallow layer (ablation)TMI regularizer · CIFAR-10 VGGSNN, FGSMaccuracy (%): 22.60 -> 25.80Improved
apply TMI temporal regularizer vs spatial MINE regularizer in ANNVGGSNN vs ANN variant · CIFAR-100, FGSMrobustness improvement: 0.43 -> 1.69Improved
add TMI regularizerAlexNet STBP · CIFAR-10, PGD TASA difference at last layerTASA difference: 0.1661 -> 0.0802Improved
add TMI regularizerSNN VGG11 T=4 · CIFAR-10, PGDaccuracy (%): 0.35 -> 26.87Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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