AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Advancing Spatiotemporal Representations in Spiking Neural Networks via Parametric Invertible Transformation

Yinsong Yan, Yujie Wu, Jibin Wu

ICLR 2026 · Architectures and efficiency › Spiking networks · approach: Training deep spiking networks · 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 learnable parametric invertible transformation (ablation)ResNet18 SNN · CIFAR10 (T=1, D=1)accuracy: 86.73% -> 91.34%Improved
add input-distribution-aware initialization (ablation)PIT ResNet18 · CIFAR10 (T=1, D=1)accuracy: 91.34% -> 92.53%Improved
add rectified surrogate gradient (ablation)PIT ResNet18 · CIFAR10 (T=1, D=1)accuracy: 92.53% -> 93.23%Improved
add PIT to LIF neuron at same latency (ablation)ResNet18 · CIFAR10accuracy: 94.22% -> 95.34%Improved
add multi-bit spikes on top of PIT (ablation)ResNet18 · CIFAR10accuracy: 95.34% -> 95.86%Improved
use fewer timesteps and more firing levels (1x4 vs 4x1) (ablation)PIT ResNet18 · CIFAR100accuracy: 76.91% -> 78.83%Improved
vary rectified gradient lambda from 0.1 to 0.001 (ablation)PIT ResNet18 · CIFAR100accuracy: 78.10% / 78.83% / 79.01%No clear effect
apply PIT to SEW ResNet34SEW ResNet34 · ImageNet-1kaccuracy: 67.04% -> 72.66%Improved
apply PIT to SEW ResNet18SEW ResNet18 · ImageNet-1kaccuracy: 63.18% -> 69.39%Improved
use PIT instead of QAT with same firing count (ablation)ResNet18 · CIFAR100accuracy: 77.91% -> 78.83%Improved
use PIT instead of I-LIF in energy termsResNet18 · CIFAR10energy (uJ): 71.94 -> 57.64 (vs LIF)Improved
apply PITVGG11 · CIFAR10-DVSaccuracy: 77.33% (TET) -> 80.50%Improved

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