Saved in:
Bibliographic Details
Main Authors: Kook, Hyunho, Yu, Byeongho, Oh, Jeong Min, Park, Eunhyeok
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.08708
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909898620534784
author Kook, Hyunho
Yu, Byeongho
Oh, Jeong Min
Park, Eunhyeok
author_facet Kook, Hyunho
Yu, Byeongho
Oh, Jeong Min
Park, Eunhyeok
contents 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
format Preprint
id arxiv_https___arxiv_org_abs_2511_08708
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stabilizing Direct Training of Spiking Neural Networks: Membrane Potential Initialization and Threshold-robust Surrogate Gradient
Kook, Hyunho
Yu, Byeongho
Oh, Jeong Min
Park, Eunhyeok
Neural and Evolutionary Computing
Computer Vision and Pattern Recognition
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
title Stabilizing Direct Training of Spiking Neural Networks: Membrane Potential Initialization and Threshold-robust Surrogate Gradient
topic Neural and Evolutionary Computing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.08708