Robustness of the Scale-free Spiking Neural Network with Small-world Property

The biological brain has the characteristics of self-adaptive, self-organizing and self-repairing. Spiking neural network (SNN) draws from the characteristics of biological brain and realizes a more advanced brain-like level. In this study, a scale-free spiking neural network (sfSNN) with small-world property is constructed, in which the lzhikevich neuron model is used as the node, the synaptic plasticity model based on the coexistence of excitatory and inhibitory synapses is used as the edge, and the scale-free network with small-world property is used as the topology. Taking the relative change rate of firing rate and the correlation between membrane potential as indexes, the robustness function of the sfSNN is analyzed. Based on the adaptive regulation of synaptic plasticity, the robustness mechanism is explored. The experimental results indicate that: (1) the sfSNN has better anti-interference function to the AC magnetic field of no more than 25 mT; (2) the sfSNN has better anti-interference function to the white Gaussian noise of no more than 10 dBW; (3) the sfSNN has better anti-injury function to the random attacks of no more than 40% injured proportion; (4) the adaptive regulation of synaptic plasticity is the intrinsic factor of the robustness function. This study lays a theoretical foundation for the engineering application of brain-like artificial intelligence.