人工神经网络并行处理的实现模型

The Implementation Models of the Parallism of Artificial Neural Network

一、人工神经网络的并行性分析人工神经网络(Artificial Neural Network,简称ANN)具有并行处理、连续计算等特点,其并行性蕴涵在神经元模型与整个网络两个逻辑层中。 1.神经元模型的并行性 ANN的基本组成单元——M-P神经元模型如图1所示。其中:(x_1,x_2,…,x_n)

It is found that the key of implementing the parallelism is to implement the parallel computation of summing the build-in weighted-product of Artificial Neuron.In this paper, the characteristics of three main kinds of Artificial Neural Network have been analysed. On the basis of introducing a two-way MAT model, another model, called one-way MAT, has been presented. They are analysed and compared on such respects as structures, mechanisms of running, overheads , advantages and limit.