基于一种自适应突触学习的动态相关系数与相位同步分析

摘要

This paper provides an adaptive synaptic learning model ,which simulates the plasticity of nerve synapses . Under the defined dynamic correlation ,it is found that non-identical neural network can be synchronized by this learning rule .It means that this learning rule has robustness .In order to express the phase synchronization of the network ,phase positioning method based on Poincare section is provided ,the position where the peak value of action potential locates is defined as Poincare section ,and the phase difference as well as network phase synchronization was are also defined .The calculation results of network phase difference show that the phase difference between any two neurons will tend to be constant along with the time ,namely ,any two neurons appear the phase synchronization ,the average phase difference of neural network tends to be constant ,and the neural network appears the phase synchronization in the whole network .%本文提出了一种自适应的突触学习模型模拟了神经突触的可塑性，通过这种学习规则在定义的动态相关系数指标下发现，可以使得一般非全同随机神经网络达到同步，表明该方法具有较好的鲁棒性。为了刻画网络在整体上的相同步提出了基于Poincare截面的相位定义法，将动作电位峰值所在的位置定义为Poincare截面，进而定义同步差，网络相位同步。网络相位差计算结果显示，任意两个神经元之间的相位差随着时间变化趋于常数，即网络中任意两个神经元出现相同步，神经网络平均相位差趋于常数，神经网络出现全局的相位同步。