Font Size: a A A

Research And Application Of BP Neural Networks Based On Genetic Algorithm

Posted on:2012-07-18Degree:MasterType:Thesis
Country:ChinaCandidate:P ZhangFull Text:PDF
GTID:2218330368478655Subject:Software engineering
Abstract/Summary:PDF Full Text Request
In recent years, artificial neural network gradually attention has been paid into the hot area of research in many fields have been involved in electronic applications such as other fields have a wide range of applications, and also continued to expand its applications. Artificial Neural Networks in Research and Application of the theory, a very important research that neural network learning algorithm, and in the algorithm is particularly important in the feed-forward neural network learning algorithm, so far no one kinds of learning feed-forward network is the ideal solution to all of the specific issues. BP neural network of the thesis as the research object, because it is the feedforward neural networks the most widely and most representative of the neural network has high research value and significance.Based on the BP neural network in-depth study, analysis of the shortcomings of BP network itself and the reasons for these shortcomings generated, a hierarchical genetic algorithm based on BP neural network optimization method using hierarchical genetic algorithms hierarchical structure to optimize the BP neural network, this hierarchical structure to retain the advantages of traditional genetic algorithm can also optimize the structure of BP neural network and the network weights. The genetic algorithm based on hierarchical structure of BP neural network optimization method hierarchical genetic algorithm and BP algorithm for global searching ability of local search combining the two algorithms in play as well as their respective advantages to overcome the shortcomings of their own .This paper proposed that a hierarchical genetic algorithm based on BP neural networkstructure optimization, the MATLAB software by Hermit polynomials and nonlinear functions to approximate and solve simple classification problems. Through analysisand comparison of experimental results demonstrate that the proposed new BP network algorithm to solve "optimal network structure" of the problem is feasible, effective, and also with the standard BP neural network algorithm comparison, analysis of the respective advantages and disadvantages. This article mainly about three aspects of the study:(1) briefly introduced the basic situation of artificial neural networks and basic theoretical knowledge. Details of the BP neural network model and BP algorithm works, analysis of the shortcomings of BP network and the causes of defects. Genetic algorithm is given the basic working principle and procedure, and analyzed the genetic algorithm in the application of BP network method.(2) shortcomings of the BP network analysis, hierarchical genetic algorithm is introduced to the needle BP network, a hierarchical genetic algorithm based on BP network structure optimization method, which can solve the structure of BP neural network optimization and easily trapped into local minima. In this paper, genetic algorithm based on hierarchical structure of BP network optimization algorithms, BP neural network structure and weights through the hierarchy genetic algorithm to optimize the hierarchical structure and hierarchical genetic algorithm using the global search ability of the BP network from local minima. And the proposed algorithm, experimental results show that the algorithm in optimizing the structure of BP network while improving convergence speed.(3) as a simulation platform with MATLAB software, applications and BP neural network method proposed in this paper a new method - genetic algorithm based on hierarchical structure of BP network optimization method, the Hermit polynomials and nonlinear function approximation and other issues the experiment, the experimental results of the analysis and comparison, a new method in this paper demonstrate the validity and rationality, but also the method to analyze the problems in order to further improve the algorithm for future Xingneng ready.
Keywords/Search Tags:BP neural networks, genetic algorithms, structural optimization, artificial neural network
PDF Full Text Request
Related items