应用新的聚类技术加速B－P网络的收敛过程

Accelerating the convergence of B-P network by using a new cluster technique

Ｂ－Ｐ网络在模式识别技术中识别效果好，应用较广泛，但通过模式样本直接对网络权训练，收敛过程太慢。就模式识别的软、硬分类问题，提出基于新的聚类方法条件下进行网络权的训练，收敛速度较快，可收事半功倍之效。

-P network has very high efficiency of recognition and has been widely used. But usually ,the pro-cess of the convergence is very slow if the weights of the network are directly trained through the patternsamples. In this paper,a procedure based on a new cluster technique is proposed to train the weights ofthe network for hard or soft classification problems of pattern recognition,and the convergence process isaccelerated.