An improved classifier based on BP and Fuzzy membership function
BP is one of most effective learning methods, implements gradient descent to reduce the error E, BP is guaranteed to converge toward local minimum in E and not necessarily to the global minimum error. An improved classifier is proposed in the paper, PSO is used to optimize the initial weight vector of BP classifier. In order to increase efficiency and adaptability of PSO algorithm, these parameters of PSO are adjusted dynamically by fuzzy membership function. The validity of the new classifier is verified by classifying iris data set.
particle swarm optimization BP algorithm fuzzy membership function
ZHANG Guoying SHA Yun
Department of information Technology, Beijing Institute of Petrochemical Technology, 102617
国际会议
北京
英文
2007-08-05（万方平台首次上网日期，不代表论文的发表时间）