AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Optimization Method for Surrogate Function in Spiking Neural Networks Based on Membrane Potential Distribution

Qi Sun, Zhen Cao, Kaige Geng, Ziyi Zhang, Biao Hou

AAAI 2026 · Architectures and efficiency › Spiking networks · approach: Spiking transformers · 1 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 membrane potential KL-divergence regularization (LDist) (ablation)MPO · CIFAR-10top-1 accuracy: 92.75% -> 94.76%Improved
add surrogate width alignment loss (Lwidth) (ablation)MPO · CIFAR-100top-1 accuracy: 70.2% -> 74.20%Improved
add dynamic adjustment strategy (DAS) (ablation)MPO · CIFAR-10top-1 accuracy: 93.65% -> 94.76%Improved
add surrogate function height (SFH) adaptation (ablation)MPO · ImageNettop-1 accuracy: 65.1% -> 65.70%Improved
add all MPO components over baseline SG training (ablation)SNN baseline · CIFAR-10top-1 accuracy: 91.55% -> 94.76%Improved
add all MPO components over baseline SG training (ablation)SNN baseline · ImageNettop-1 accuracy: 63.3% -> 65.70%Improved
add discretization adjustment strategy (ablation)MPO · CIFAR-10 accuracy gap between T=20 and T=8accuracy gap: 11.3% -> 6.9%Improved
add discretization adjustment strategy (ablation)MPO · CIFAR-10, T=8accuracy: 80.9% -> 85.8%Improved
add MPO training to SNN baselineResNet-18 SNN · Power consumptionspike density: 0.65 -> 0.47Improved
add MPO training to SNN baselineVGG-16 SNN · Power consumptionSynOps: 23.0M -> 19.5MImproved
add MPO training under noiseSNN baseline · CIFAR-10 Gaussian noise sigma=0.2accuracy: MPO retains 91.97% at sigma=0.2Improved

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