AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Training Deep Spiking Neural Networks without Normalization

Xinyu Shi, Zhaofei Yu

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

Read the paper

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 SpikeInit weight initialization (firing-rate response curve) (ablation)normalization-free SNN · CIFAR-10 VGG-11 / ResNet-18accuracy: fails to converge -> 95.29 (VGG-11); fails -> 95.53 (ResNet-18)Improved
add surrogate gradient shape parameter initialization (ablation)SpikeInit · CIFAR-10 VGG-11accuracy: 78.96 -> 95.29Improved
add surrogate gradient shape parameter initialization (ablation)SpikeInit · CIFAR-10 ResNet-18accuracy: 91.05 -> 95.53Improved
add adaptive surrogate gradient (ablation)SpikeInit · CIFAR-10 VGG-11accuracy: 94.43 -> 95.29Improved
add adaptive surrogate gradient (ablation)SpikeInit · CIFAR-10 ResNet-18accuracy: 94.91 -> 95.53Improved
replace Micheli et al. initialization with SpikeInitnormalization-free SNN · CIFAR-10 VGG-11/VGG-16accuracy: fails to converge -> 95.29 (VGG-11), 95.77 (VGG-16)Improved
replace Micheli et al. initialization with SpikeInitMS ResNet-34 · CIFAR-10accuracy: 91.83 -> 96.20Improved
remove normalization and use SpikeInit instead of TEBNMS ResNet-34 · CIFAR-10accuracy: 95.87 -> 96.20Improved
remove normalization and use SpikeInit instead of TEBNMS ResNet-34 · ImageNet (4 time steps)accuracy: 68.28 -> 68.81Improved
replace LayerNorm with SpikeInit in spiking transformerspiking transformer 4-256 · IWSLT17 German-EnglishBLEU: 29.50 -> 29.98Improved
replace BatchNorm (tdBN) with SpikeInit in spiking transformerspiking transformer 4-256 · IWSLT17 German-EnglishBLEU: 28.14 -> 29.98Improved
increase depth to 1000 layers with SpikeInit vs tdBN baselineMS-ResNet · CIFAR-10 depth scalingaccuracyImproved

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