Dissertation > Excellent graduate degree dissertation topics show

The Application and Research of RBF Neural Network Optimized by Cultural Algorithms

Author: WeiXiu
Tutor: ZhangXueYing
School: Taiyuan University of Technology
Course: Communication and Information System
Keywords: Speech Recognition RBF neural network Cultural Algorithm Niche algorithm Dynamic accept the function
CLC: TP183
Type: Master's thesis
Year: 2011
Downloads: 75
Quote: 0
Read: Download Dissertation

Abstract

With the ever-accelerating pace of life, people are more eager to be able to easily and quickly communicate directly with the computer. Voice as an ideal human-computer interaction, making the field of research priorities. Speech recognition technology carries this mission came into being. At this stage, the speech recognition technology is still mainly based on the study of linear systems, which fetters the further development of this technology, the researchers realized that in order to achieve substantive progress without delay the introduction of nonlinear methods. Since the 1980s, artificial neural networks for solving nonlinear theory of nonlinear problems become a research hotspot. RBF as a new effective feedforward neural network, with its simple structure, fast training, classification performance, generalization ability, etc., in the field of speech recognition to get more and more attention and application. RBF topology of each parameter is determined dynamically during training, the network has better adaptability. However, the performance of the network node based on the implicit function is very sensitive to the center of the selection. As a recent cultural algorithm optimization algorithm has good global search performance. Content of this study was to use cultural algorithm to optimize the RBF neural network hidden node basis function centers to improve network performance, and apply speech recognition systems. Research work mainly includes the following points: 1, the introduction of a new kind of cultural performance with global search algorithm framework model. Since the algorithm is still in its infancy, has not yet formed the inherent mode model depends on the setting of specific issues, so I need for this study for solving nonlinear unconstrained optimization problems, the design of the cultural algorithm framework component. Where population space using genetic algorithms, and through simulation experiments on cultural algorithm and genetic algorithm when used alone, compare the performance of the algorithm. 2, the design niche dynamic receiver function. Culture algorithm accepts a function that affect the performance of the key cultural algorithms directly affects convergence of the algorithm and operational efficiency. Therefore, this paper introduces niche algorithm dynamically select the size of each niche mechanism after each iteration according to the actual situation of the individual groups adaptively adjust the number of individuals receiving excellent proposed niche dynamic acceptance function. And through simulation experiments and the use of several existing culture that accepts a function to compare performance of the algorithm. 3, with cultural training RBF neural network algorithm to determine the center of its base function to optimize the RBF network, and applied to speech recognition systems.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory > Artificial Neural Networks and Computing
© 2012 www.DissertationTopic.Net Mobile