Deferring the Learning for Better Generalization in Radial Basis Neural Networks

摘要: The level of generalization neural networks is heavily dependent on the quality training data. That is, some patterns can be redundant or irrelevant. It has been shown that with careful dynamic selection patterns, better performance may obtained. Nevertheless, carried out independently novel to approximated. In this paper, we present a learning method automatically selects most appropriate new sample predicted. proposed applied Radial Basis Neural Networks, whose capability usually very poor. strategy slows down response network in generalisation phase. However, does not introduces significance limitation application because fast Networks.