【摘要】 本文利用Ｋｏｈｏｎｅｎ侧抑制自组织神经网络在特征映射过程中，其样本内部拓扑性保持一致的特性，对三维多目标识别中的训练样本进行了合理优化选择，并用增量算法进行验证．结果表明：由减少的训练样本训练网络后，网络的样本识别率仍能达到较高水平，从而实现了减少训练样本和网络训练时间、而神经网络的识别率及泛化性不受影响的良好效果更多还原

【Abstract】 Kohonen self organized neural network was used in optimal selection of training set in pattern recognition.The increment algorithm was used to test the results.Computer simulation indicated that although fewer training samples were adopted,the recognition ratio and the generalization of this neural network was high enough.更多还原