一种带神经网络的并行VLSI体系结构的算法

Algorithm and Implementation of Parallel VLSI Architectures Using Neural Networks

本文提出一种适合完成复杂信息处理任务的并行VLSI结构的算法。该算法基于复现神经网络,完成自动过程分配和次优路线选取,并利用确定网络内使用的能量函数极小化实现处理程序和路线选取,从而设计出数据流图的有效结构及数据通路。

This paper proposes an algorithm of parallel/pipelined VLSI architectures suitable for high complicated information processing.The algorithm can accomplish the automatic process allocation and the sub-optimum routing using recurrent neural network.The report shows that the process scheduling and routing can be accomplished by well determining an energy function minimization used in the RNN and that the efficient architecture of a data flow graph can be designed.