arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.07037cs.NEcs.AIcs.CVcs.ET
收藏

基于随机脉冲神经元的SNN可以具有内在贝叶斯性

Stochastic Spiking Neuron Based SNN Can be Inherently Bayesian

The Hong Kong University of Science and Technology (GuangZhou)(香港科技大学(广州))

Huannan Zheng, Jingli Liu, Kezhou Yang

AI总结:

本文提出了一种结合设备随机性和随机阈值神经元的脉冲贝叶斯神经网络,通过利用噪声作为贝叶斯资源,实现了高准确率和鲁棒性,同时提升了训练效率。

摘要:

生物神经系统中的不确定性似乎在计算上是有益而非有害的。然而,在神经形态计算系统中,设备的变异性往往限制性能,包括准确性和效率。在本工作中,我们提出了一种脉冲贝叶斯神经网络(SBNN)框架,该框架统一了内在设备随机性(基于磁隧道结)的动态模型和随机阈值神经元,以利用噪声作为功能贝叶斯资源。实验表明,SBNN在MNIST上达到99.16%的高准确率,在CIFAR10上达到94.84%的准确率,使用8位精度。同时,速率估计方法提供了约20倍的训练加速。此外,SBNN表现出优越的鲁棒性,在突触权重噪声下比标准脉冲神经网络提高了67%的准确率,在输入噪声下提高了12%。关键的是,硬件验证确认物理设备实现相比算法模型导致不可见的准确性和校准损失。将设备随机性转化为神经元不确定性为在不确定性下实现紧凑、节能的神经形态计算提供了一条途径。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑