An Online Supervised Learning Algorithm Based on Spike Train Kernel for Spiking Neurons

Liu Yang, Xianghong Lin, Yuan Zhao, Xiangwen Wang, Xianrun He, Yaohui Zou

Abstract: Spiking neural networks, heralded as the third generation of artificial neural networks, spiking neurons utilize a precision timing encoding strategy for the processing of neural information. Neuroscience studies show that online learning for spiking neurons is more consistent with biological basis. Using the spike train kernel definition as a foundation, the paper introduces an online supervised learning algorithm tailored for spiking neurons incorporating temporal encoding. The algorithm was employed in various spike train learning tasks, analyzing the impact of different learning parameters on the algorithm’s learning performance and comparing the learning performance with offline learning algorithms. The experimental results demonstrate that our presented method achieves a higher level of accuracy; it can proficiently address complex spatio-temporal spike pattern learning challenges.