基于改进遗传算法进化神经网络.pdf

（ ） — Journai of Southeast University Engiish Edition Dec . 2002 Voi . 18 No .4 ISSN 1003 7985 Evolving Neural Networks Using an lmproved Genetic Algorithm ! ， Wen Xiuian1 2 !! Song Aiguo1 Duan Jianghai1 Wang Yiging1 1 （ ， ， ， ） Department of Instrument Science and Technoiogy Southeast University Nanjing 210096 China 2 （ ， ， ， ） Mechanicai Engineering Coiiege Inner Mongoiia Poiytechnic University Huhot 010062 China Abstract： （ ） A novei reai-coded improved genetic aigorithm GA of training feed-forward neurai network is proposed to reaiize noniinear system forecast . The improved GA empioys a generation-aiternation modei based the minimai generation gap （ ） （ ） ， MGP and biend crossover operators BLX-! . Compared with traditionai GA impiemented in binary number the ， ’ processing time of the improved GA is faster because coding and decoding are unnecessary . In addition it needn t set parameters such as the probabiiity vaiue of crossover and mutation by experiences . Therefore，