Stabilizing Direct Training of Spiking Neural Networks: Membrane Potential Initialization and Threshold-robust Surrogate Gradient

Hyunho Kook, Byeongho Yu, Jeong Min Oh, Eunhyeok Park
11/11/2025
cs.NEcs.CV

Official code: kookhh0827/SNN-MP-Init-TRSG (3 stars, last commit 2026-04-03).

Abstract

Recent advancements in the direct training of Spiking Neural Networks (SNNs) have demonstrated high-quality outputs even at early timesteps, paving the way for novel energy-efficient AI paradigms. However, the inherent non-linearity and temporal dependencies in SNNs introduce persistent challenges, such as temporal covariate shift (TCS) and unstable gradient flow with learnable neuron thresholds. In this paper, we present two key innovations: MP-Init (Membrane Potential Initialization) and TrSG (Threshold-robust Surrogate Gradient). MP-Init addresses TCS by aligning the initial membrane potential with its stationary distribution, while TrSG stabilizes gradient flow with respect to threshold voltage during training. Extensive experiments validate our approach, achieving state-of-the-art accuracy on both static and dynamic image datasets. The code is available at: https://github.com/kookhh0827/SNN-MP-Init-TRSG

View on arXivView PDF

Code Implementations(1)

kookhh0827/SNN-MP-Init-TRSGOfficial100%

[WACV2026] Stabilizing Direct Training of Spiking Neural Networks: Membrane Potential Initialization and Threshold-robust Surrogate Gradient

30Shell, PythonNov 9, 20256 months ago

Similar Papers

Robust Stable Spiking Neural Networks

85% match1 repo82024

SSF: Accelerating Training of Spiking Neural Networks with Stabilized Spiking Flow

82% match2023

Equilibrium Propagation: Bridging the Gap Between Energy-Based Models and Backpropagation

82% match1 repo1182016

Stabilizing RNN Gradients through Pre-training

82% match2023

Gradient descent at the Edge of Stability: free energy model and kinetic description of the two-layer network

82% match2026

InfLoR-SNN: Reducing Information Loss for Spiking Neural Networks

81% match2023

Cite this paper

@article{kook2025stabilizing,
  title  = {Stabilizing Direct Training of Spiking Neural Networks: Membrane Potential Initialization and Threshold-robust Surrogate Gradient},
  author = {Hyunho Kook and Byeongho Yu and Jeong Min Oh and Eunhyeok Park},
  year   = {2025},
  eprint = {2511.08708},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2511.08708v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2511.08708.svg)](https://codeofpaper.com/paper/2511.08708)

Discussion