Stochastic Spiking Neuron Based SNN Can be Inherently Bayesian

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zheng, Huannan, Liu, Jingli, Yang, Kezhou
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917254842548224
author Zheng, Huannan
Liu, Jingli
Yang, Kezhou
author_facet Zheng, Huannan
Liu, Jingli
Yang, Kezhou
contents Uncertainty in biological neural systems appears to be computationally beneficial rather than detrimental. However, in neuromorphic computing systems, device variability often limits performance, including accuracy and efficiency. In this work, we propose a spiking Bayesian neural network (SBNN) framework that unifies the dynamic models of intrinsic device stochasticity (based on Magnetic Tunnel Junctions) and stochastic threshold neurons to leverage noise as a functional Bayesian resource. Experiments demonstrate that SBNN achieves high accuracy (99.16% on MNIST, 94.84% on CIFAR10) with 8-bit precision. Meanwhile rate estimation method provides a ~20-fold training speedup. Furthermore, SBNN exhibits superior robustness, showing a 67% accuracy improvement under synaptic weight noise and 12% under input noise compared to standard spiking neural networks. Crucially, hardware validation confirms that physical device implementation causes invisible accuracy and calibration loss compared to the algorithmic model. Converting device stochasticity into neuronal uncertainty offers a route to compact, energy-efficient neuromorphic computing under uncertainty.
format Preprint
id arxiv_https___arxiv_org_abs_2602_07037
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stochastic Spiking Neuron Based SNN Can be Inherently Bayesian
Zheng, Huannan
Liu, Jingli
Yang, Kezhou
Neural and Evolutionary Computing
Artificial Intelligence
Computer Vision and Pattern Recognition
Emerging Technologies
Uncertainty in biological neural systems appears to be computationally beneficial rather than detrimental. However, in neuromorphic computing systems, device variability often limits performance, including accuracy and efficiency. In this work, we propose a spiking Bayesian neural network (SBNN) framework that unifies the dynamic models of intrinsic device stochasticity (based on Magnetic Tunnel Junctions) and stochastic threshold neurons to leverage noise as a functional Bayesian resource. Experiments demonstrate that SBNN achieves high accuracy (99.16% on MNIST, 94.84% on CIFAR10) with 8-bit precision. Meanwhile rate estimation method provides a ~20-fold training speedup. Furthermore, SBNN exhibits superior robustness, showing a 67% accuracy improvement under synaptic weight noise and 12% under input noise compared to standard spiking neural networks. Crucially, hardware validation confirms that physical device implementation causes invisible accuracy and calibration loss compared to the algorithmic model. Converting device stochasticity into neuronal uncertainty offers a route to compact, energy-efficient neuromorphic computing under uncertainty.
title Stochastic Spiking Neuron Based SNN Can be Inherently Bayesian
topic Neural and Evolutionary Computing
Artificial Intelligence
Computer Vision and Pattern Recognition
Emerging Technologies
url https://arxiv.org/abs/2602.07037