A STDP Rules-Based Spiking Neural Network Implementation for Image Recognition

With the progress and development of artificial intelligence, the concept of neuromorphic computing has been widely observed since its proposed. The biological neural system can use sparse electrical spikes for information transfer, has powerful memory-learning functions, and perform complex tasks with extremely low power consumption. As the third generation neural network, spiking neural network has higher level of bionic characteristics than the traditional neural network. It can simulate the information processing mechanism of human brain to a greater extent, which is the focus of the research of brain-like computing. In this work, a two-layer SNN composed of Leaky-Integrate-Fire neurons was implemented in the BindsNET simulation environment. The network was learned and trained by the time-dependent plasticity of spike rule, and achieves 92% recognition accuracy on the MNIST handwritten digit set.