基于 Simulink 平台的脉冲神经网络前向传播模型的建立与仿真

摘要

脉冲神经网络（SNN）被誉为第三代神经网络，近年来受到许多学者的关注，其优势已经在模式识别、计算机视觉等诸多领域得到了发挥。脉冲神经网络的硬件化是实现其强大计算能力的重要途径，而使用相关的软件对硬件系统进行建模和仿真，是复杂硬件系统设计的首要问题。为了获得合理的硬件设计方案，在 Simulink 平台上，首先建立脉冲神经网络突触的模型，并得到较为理想的突触响应曲线。在此基础上，建立一个完整的脉冲神经网络前向传播模型。仿真结果显示，通过训练可以解决传统的异或问题。%Spiking neural network (SNN)is praised as the third generation of the neural networks,and has attracted concerns from many scholars in recent years.Its advantages have been brought into play in a couple of fields including pattern recognition and computer vision, etc.The hardware realisation of spiking neural network is an important approach for implementing its powerful computation ability,while modelling and simulating the hardware system with the help of related software is a primary issue in complex hardware system design.In order to get reasonable hardware design scheme,on Simulink platform we first build the synapse model of SNN,and obtain an ideal response curve of synapse.Based on this,we build an integrated forward propagation model of spiking neural network.Simulation results show that the traditional XOR problem can be resolved through training.