摘要： 针对复杂的模式识别问题，提出了一种用串-并行混合结构的多种神经网络模型进行求解的方法。首先用ART网络对训练集中的样本进行粗分类，以减小训练集的样本规模，然后用多个BP网络并行地对小训练集进行训练。用于汉字识别问题，取得了很好的效果。

关键词: 模式识别, 串-并行结构, 神经网络, 多模型

Abstract: In order to solve complicated problems of pattern recognition, a method of multi-model of artificial neural networks with serial-parallel structure is proposed.Firstly, sample set is roughly classified using ART to reduce the scale of samples ，in training set,and then all small training sets is trained using parallel BP. Applying to Chinese characters recognition, this method is proved to achieve better result.

中图分类号: 

• TP183