人工神经网络的局域性对网络训练速度的影响

从网络结构入手，提出了网络局域性的概念。作为网络结构的一种定量描述，探讨了网络结构与训练速度、预测精度间的对应关系。结果表明网络的训练速度随局域性的增加而增加，网络的预测精神在局域性０．５５附近达到最高，任何偏离都会导致网络预测精度的下降。为在生化过程具体应用中选择合理的神经网络类型提供了理论依据。

The Effect of the Artificial Neural Network's Provincialism on Its Training Rate

There is on theory that gives the method to select the special kind of artificial neural network (ANN) in the biochemical engineering. The idea of provincialism of artificial neural network is introduced to describe the structural difference of various kinds of ANN. Based on the typical equation in the biochemical engineering, we discussed the effects of ANN's topological structure on its training rate, and on its prediction error in this paper. The results show an increase of the network's training rate as its provincialism increases from 0 to 1. When provincialism is about 0.55, the prediction error falls to the lowest, Any variance from this value will lead to an increase in prediction error. Then, the research provides a method by which the most optima network can be selected in biocheAncal engineering modeling field.