不同聚类系数无标度脉冲神经网络动态特性对比分析

COMPARATIVE ANALYSIS OF DYNAMIC CHARACTERISTICS OF SCALE-FREE SPIKING NEURAL NETWORKS WITH DIFFERENT CLUSTERING COEFFICIENTS

拓扑结构可以影响脉冲神经网络的功能。聚类系数是反映网络连通性的重要拓扑性质之一,基于不同聚类系数拓扑进行脉冲神经网络的研究对于进一步理解网络拓扑性质对网络功能影响具有重要意义。以Izhikevich神经元模型为节点,突触可塑性为边,分别构建高、低聚类系数无标度拓扑的脉冲神经网络,评估并对比两种网络的动态特性。实验结果表明,两种网络动态特性随时间均先剧烈变化而后逐渐趋于稳定;在网络连接强度、局部和全局信息传输效率、小世界属性方面,高聚类均比低聚类无标度脉冲神经网络更具优势。

The structure of topology can affect the function of the spiking neural network.Clustering coefficient is one of the important topological properties reflecting network connectivity,and the research of spiking neural network based on the topology with different clustering coefficients is of great significance to understand the influence of network topological properties on network functions.This paper used the Izhikevich neuron model as a node and synaptic plasticity as an edge,constructed scale-free spiking neural networks(SFSNN)with high and low clustering coefficient topologies,and evaluated and compared the dynamic characteristics of the two networks.The experimental results show that the dynamic characteristics of the two kinds of networks first change sharply with time,and then gradually tend to be stable.The SFSNN with high clustering coefficient has better performance than the SFSNN with low clustering coefficient in terms of network connection strength,local and global information transmission efficiency and small-world property.