Improved Bp Neural Network In Pattern Recognition Applications And Research

Artificial Neural Network has great nonlinear mapping ability and it has been widely used in many fields such as Pattern Recognition, artificial controlling, image processing, time series analyst. Having so many special functions like self-organization, self-studying and fault-tolerant, it has particular effects when processing complicated nonlinear problems, and it has been the hotspot at home and abroad.This thesis first introduces the development of pattern recognition using Artificial Neural Network in details to get the superiority of Neural Network Recognition over traditional pattern recognition methods.Then on the basis of existing Neural Network Recognition system, the thesis introduces the BP Neural Network, which is most widely used in the system. By studying the critical technology and algorithm and focusing on the disadvantage of BP Algorithm, we bring out the heuristic improvement of BP Algorithm. Meanwhile, by analyzing and studying the Genetic BP model and its algorithm, we introduce several common types of improvements aiming at the main defection of Genetic Algorithm.This thesis improves the BP Algorithm combined with Genetic Algorithm and establishes the modified GA-BP Genetic Algorithm Model.Finally, by taking advantage of the two programming languages, VC++and MATLAB. this thesis designs the Genetic BP Neural Network Recognition System based on the proposed GA-BP algorithm and puts it into operation.The Recognition System has been used to carry on training and testing of two actual pattern recognition problems. The experimental results indicate that the GA-BP Genetic Algorithm can deal with the problem of the network training failure defection because of the partial minimum during the network training, and can improve the generalizing ability.