论文部分内容阅读
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，探讨了网络结构与网络训练速度、预测精度间的对应关系．结果表明网络的训练这度随局域性的增加而增加，网络的预测精度在局域性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．
Starting from the network structure, this paper proposes the concept of network localization as a quantitative description of network structure, and discusses the correspondence between network structure and network training speed and prediction accuracy. The results show that the training degree of the network increases with the increase of the locality. The prediction accuracy of the network reaches the highest around the local area of 0.55, and any deviation will lead to the decline of the network prediction accuracy. Which provides a theoretical basis for choosing a reasonable type of neural network in the specific application of biochemical process.