新型BP网络用于非线性化学体系的研究

A New BP Neural Network Applied to the Study of Chemical Non-linear Systems

提出了一种杂化ＢＰ网络模型及新型变步长算法，使常规ＢＰ算法的收敛平稳性及收敛速度获得了改善，对两类非线性化学体系的实例研究表明该网络较多元线性回归及常规ＢＰ网络具有更好的学习与预测效果。

A new hybrid BP neural network model and a novel varying steplength algorithm was presented and used to the study of the real chemical nonlinear systems in this paper. The given model and algorithm had a good improvements in convengent stability and rate compared to MLR and convensional BP neural networks.