Fractional Linear Neural Network Based On Genetic Algorithm Design And Applications

Artificial Neural networks (ANN), characterized with highly parallel &non-linear feature and distributed memory, is a new information-processing system based on the initial acquaintance with the organization structure of human brains as well as activity mechanism. Fractional neural network (FNN), which is proposed by Yang Guowei, is a new neural network which has a merit of good approximation capability. However, it is still a difficult problem concerning with the optimization design of FNN.Based on the National Natural Science Foundation project on"Artificial brain information processing new neural network model", No.60673101 for the research task, the thesis makes full use of the respective advantages of genetic algorithm (GA) and the BP algorithm for a FNN, combining the GA with BP algorithm and proposing a GA-based FNN optimization design algorithm and method. Take the case of predicting rainfall of Daqing, the paper proving that GA-based FNN optimization design algorithm is feasible and effective by simulation results.