Study of Small-world Spiking Neural Network Based on Anti-interference Ability against Magnetic Field

Abstract

The study of robustness contributes to improving the neural information processing ability of brain-like models. However, the effect of topological structures of brain-like models on anti-interference ability has not been clarified. In this study, integrating the Izhikevich neuron model, the synaptic plasticity model, and the small-world network, a small-world spiking neural network (SWSNN) is constructed. Then, the anti-interference ability of the SWSNN is assessed based on two anti-interference indicators under AC magnetic field. Finally, the anti-interference performance of SNNs with different topologies is compared. The simulation results consistently verify that: (i) The SWSNN has anti-interference ability, and the intrinsic factor of anti-interference ability is the synaptic plasticity. (ii) In terms of anti-interference performance, the SWSNN outperforms the scale-free SNN, which hints that topological structure affects the anti-interference ability at the level of performance.