AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Optimized Potential Initialization for Low-Latency Spiking Neural Networks

Tong Bu, Jianhao Ding, Zhaofei Yu, Tiejun Huang

AAAI 2022 · Architectures and efficiency › Spiking networks · approach: Training deep spiking networks · 91 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
Initialize membrane potentials to half the firing threshold rather than zero (ablation)Converted SNNs · CIFAR-10, CIFAR-100, and ImageNet; low time-stepsaccuracyImproved
Use half-threshold initial potentials rather than zero in RMP (ablation)RMP, ResNet-20 · CIFAR-10, T=64accuracy: 67% -> 87%Improved
Use half-threshold initial potentials rather than zero in RNL (ablation)RNL, ResNet-20 · CIFAR-10, T=64accuracy: 45% -> 82%Improved
Use half-threshold initial potentials rather than zero in RNL (ablation)RNL, ResNet-20 · CIFAR-10, T=32accuracy: 25% -> 70%Improved
Use zero rather than nonzero initial membrane potentials (ablation)Converted SNNs · CIFAR-10 and CIFAR-100, low latencyaccuracyHurt
Use uniform-random or Gaussian-random initial potentials rather than optimal half-threshold potentials (ablation)Converted SNNs · CIFAR-10 and CIFAR-100, low latencyaccuracyHurt
Add trainable clipping bounds to the source ANN and use them as SNN thresholdsANN-to-SNN conversion · CIFAR-10, CIFAR-100, and ImageNetaccuracyImproved

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