提出了一种新的自组织神经网络模型 ,这种网络克服了Kohonen自组织神经网络不能直接处理划分和布局问题的面积约束、邻域确定困难和学习速度慢等缺点 .还将这一神经网络模型应用于以芯片间连线代价最小和时钟周期最小为优化目标的、以面积和时延为约束条件的性能驱动的MCM系统划分中 .算法不仅学习速度快 ,能够处理大规模电路划分问题 ,而且具有整体优化效果 .本算法已用VisualC + +编程实现 ,实验结果表明 ,这是一种有效的方法 A new model of self-organizing neural network is proposed, which overcomes the shortcomings of Kohonen’s self-organizing neural network that can not deal directly with the partition and layout problems, the difficulty of determining the neighborhood and the slow learning speed etc. This neural network The model is applied to the performance-driven MCM system partitioned by the area and the delay as the optimization objective, which has the lowest cost and the smallest clock cycle, and the algorithm not only learns fast, but also can deal with the large-scale circuit partitioning problem, But also has the overall optimization effect.The algorithm has been programmed with VisualC + +, the experimental results show that this is an effective method