基于神经网络的并行布局算法

Parallel placement algorithm based on neural network

根据神经网络串行布局算法 ,进一步研究并提出了与布局问题相适应的神经网络并行计算模型及并行布局算法 ,以负载平衡及系统通信开销最小为目标 .理论分析结果表明 ,运用该算法能达到此目标 ,且具有布局质量高、收敛快、计算简单等特点 。

This article studies the application of neural network theory in automatic placement. According to the serial placement algorithm based on neural network, it describes and presents further neural network parallel calculation model and parallel placing algorithm suitable to the placement problem to meet the goal of load balance and communication overhead minimum. The results of theoretical analysis show that the algorithm meets the goal. The performance of the method has high placing quality, quick convergence, simple calculation, fitting to the neural network parallel placement.