Integration of Learning Algorithm on Fuzzy Min-Max Neural Networks

• 作 者: 胡静 罗宜元
• 机 构: School of Electronic Information Engineering,Shanghai Dianji University,
• 出 处: Journal of Shanghai Jiaotong University(Science)
• 发表日期: 2017年06期
• 关键词: clustering similarity superiority Simpson supervised boxes membership overlapped

An integrated fuzzy min-max neural network(IFMMNN) is developed to avoid the classification result influenced by the input sequence of training samples, and the learning algorithm can be used as pure clustering,pure classification, or a hybrid clustering classification. Three experiments are designed to realize the aim. The serial input of samples is changed to parallel input, and the fuzzy membership function is substituted by similarity matrix. The experimental results show its superiority in contrast with the original method proposed by Simpson.

An integrated fuzzy min-max neural network (IFMMNN) is developed to avoid the classification result influenced by the input sequence of training samples, and the learning algorithm can be used as pure clustering, pure classification, or a hybrid clustering classification. Three experiments are designed to realize the aim. The serial input of samples is changed to parallel input, and the fuzzy membership function is substituted by similarity matrix. The experimental results show its superiority in contrast with the original method proposed by Simpson.