基于Izhikevich模型的神经网络放电特性数值模拟  

Numerical Simulation on Discharge Characteristics of Neural Network Based on Izhikevich Model

作　　者：王超 WANG Chao(Key Laboratory of Optoelectronic Technology and Intelligent Control Ministry of Education,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学光电技术与智能控制教育部重点实验室,甘肃兰州730070

出　　处：《工业技术创新》2020年第2期33-37,共5页Industrial Technology Innovation

摘　　要：利用数值模拟方法研究神经网络的放电特性,为人工智能等领域的研究提供理论支持。以Izhikevich模型为节点,用化学突触连接各神经元,搭建神经网络,分析兴奋性突触和抑制性突触在神经网络中的不同效应,并通过调试化学突触的参数,实现峰放电和簇放电在神经网络中的正常传递。研究发现:当抑制性突触电导为0.5 mS/cm^2时,突触后神经元动作电位传递受到抑制;当兴奋性突触电导为0.5 mS/cm^2时,神经元动作电位传递正常;当兴奋性突触电导为3 mS/cm^2时,突触后神经元动作电位由峰放电变为簇放电,发生畸变;选用参数合适的兴奋性突触可以使神经网络实现峰放电和簇放电的正常传递,但化学突触存在时间滞后性。The discharge characteristics of the neural network are studied by numerical simulation,to provide theoretical support for the research of fields such as artificial intelligence.Taking Izhikevich model as the node,chemical synapses are used to connect neurons to build the neural network,and different effects of excitatory synapses and inhibitory synapses in the neural network are analyzed.By adjusting the parameters of chemical synapses,the transmissions of spiking discharge and clustering discharge in the neural network are realized.The study shows that,when the inhibitory synaptic conductance is 0.5 mS/cm^2,the transmission of the postsynaptic action potential is inhibited;when the excitatory synapse conductance is 0.5 mS/cm^2,the action potential is transferred normally;when the excitatory synapse conductance is 3 mS/cm^2,the postsynaptic action potential changes from spiking discharge to clustering discharge,resulting in distortion.Selecting excitatory synapses with appropriate parameters make the neural network achieve the transmission of spiking discharge and clustering discharge,but the chemical synapses show the characteristics of time lag.

关 键 词：Izhikevich模型 神经网络 数值模拟 放电特性 化学突触 时间滞后性 

分 类 号：R322.8[医药卫生—人体解剖和组织胚胎学] Q811.2[医药卫生—基础医学]