Research On Approximation Capability Of RBF Neural Networks

As a class of intelligent computational methods, neural network has been applied in many areas, such as engineering, computer science, physics, biology, economy and management, etc. It can be effectively used on data excavation, cluster analysis, intelligent control, pattern recognition and optimality calculation, etc. This leads to many fundamentals research on the architecture, convergence, approximation ability, and generalization capability of neural network. The aim of this paper is to study the approximation ability of the RBF neural network in sense of IP (K).There are three parts in this article: The first chapter of this paper introduces the foundation of neural network. The second chapter introduces the architecture of the RBF neural network, the ways of training, and some fundamental research on it. In the third part of this paper, we proved one of the lemmas for our results in research of the approximation capability of RBF neural networks, which relaxes the requirement for function g.