An Adaptive Structure Learning Algorithm for Multi-Layer Spiking Neural Networks

Due to the complex multi-layer structure and implicit nonlinear mechanism, the formulation of efficient learning methods for spiking neural networks with dynamically adaptive structure is difficult. This paper presents an adaptive structure learning algorithm for multi-layer spiking neural networks, in which the synaptic weights are modified according to the supervised learning algorithm based on inner product of spike sequences. The main contribution of this work lies in the fact that the proposed algorithm is able to dynamically prune neurons of the hidden layer. The proposed algorithm is successfully applied to learn spikes sequences. The experimental results verify the effectiveness of the adaptive structure learning algorithm in multi-layer spiking neural networks. Moreover, the adaptive structure networks achieve better performance in the spike train leaning task than the fixed structure networks.