用噪声当资源，让脉冲神经网络天然具备贝叶斯推理能力
Stochastic Spiking Neuron Based SNN Can be Inherently Bayesian
- 将器件随机性与脉冲神经元阈值随机性统一建模，把噪声变作贝叶斯计算资源
- 在MNIST上达99.16%准确率，CIFAR10上达94.84%，8比特精度下仍高效
- 对突触和输入噪声鲁棒性强，适合低功耗、抗干扰的神经形态硬件部署
生物神经系统的不确定性似乎具有计算优势而非缺陷。然而，在神经形态计算系统中，器件变异常限制性能，包括准确率和效率。本文提出一种脉冲贝叶斯神经网络（SBNN）框架，统一建模基于磁隧道结的内在器件随机性与随机阈值脉冲神经元，将噪声作为功能性贝叶斯资源加以利用。实验表明，SBNN在MNIST上达到99.16%准确率，CIFAR10上达94.84%，且仅需8比特精度。同时，速率估计方法使训练速度提升约20倍。此外，SBNN展现出更强鲁棒性：在突触权重噪声下准确率提升67%，在输入噪声下提升12%，优于标准脉冲神经网络。关键的是，硬件验证证实物理器件实现导致与算法模型相比无可见准确率损失和校准损耗。将器件随机性转化为神经元不确定性，为在不确定性环境下实现紧凑、低功耗神经形态计算提供了路径。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。