The Research On The Spiking Neural Network And Its Applications

Spike neuron is the third generation of artificial neuron. With biology neuron dynamic firing property, it can deal with time pattern problems. So it is an important task to research on the Spike neuron in the neural network field. While the research on the learning pattern and algorithm of Spiking neural network can help to find the computing properties of Spike neuron, which also benefits solve the actual problems. In this paper, several common Spike models, especially the Spike response model (SRM), is analyzed first, then a Spiking neural network with particle swarm optimization is proposed, which is used to solve pattern recognition and bioinformatics problems. Protein secondary structure prediction is one of the important problems of bioinformatics. It can obtain function properties of protein through predicting secondary structure. So the problem of protein secondary structure prediction is also discussed in this paper, two encoding methods of amino acid are applied on the single and multi-level Spiking neural networks. The main contributions of this paper are as follows:(1) A Spiking neural network model with particle swarm optimizationThe SRM model, the encoding property of Spike neuron and SpikeProp algorithm are analyzed first. The SpikeProp uses gradient information to adjust parameters, which may be trapped in local optima, so the particle swarm optimization (PSO) based on global optimazation is applied to Spiking neural network. Unlike backpropagation-based SpikeProp algorithm, the particle swarm optimization does not require gradient information and can provide a global stochastic optimal search for both positive and negative weights. Moreover the network only uses one synaptic connection between two neurons characterized by a weight and a delay value, therefore, the network is simplified. The PSO-based learning model sets the foundation for applications.(2) Solution of pattern recognize problemsThe learning model based on PSO is used in the Spiking neural network for solving the XOR and international standard classify IRIS problem. Compared to SpikeProp algorithm, the PSO model can advance the learning speed and improve the classify accuracy.(3) Study on the protein secondary structure prediction with Spiking neural network...