基于改进神经网络的实验方案优选系统  

The Optimal Selection System of Experiment ConditionsBased on the Improved ANN

作　　者：潘丹[1] 罗干英[1] 黄茜[1] 肖诗铁[1] 

机构地区：[1]华南理工大学电子与通信工程系

出　　处：《暨南大学学报（自然科学与医学版）》1998年第1期81-87,共7页Journal of Jinan University(Natural Science & Medicine Edition)

摘　　要：构造了一类新的高效分段活化函数，很好地解决了ＢＰ算法学习收敛速度慢的问题，并提出了一种自适应调整网络参数的新算法，从而大大提高了算法的学习效率和综合性能．In this paper,not only a kind of new activation functions,called segment activation function(SAF),but also the improved back propagation algorithm in which networks parameters can be adaptively adjusted is proposed to solve two key problems:slow convergence and low learning efficiency which exist in the conventional BP ANN and restrict its applications,so the learning efficiency and comprehensive properties are greatly improved.Moreover,the procedure of modeling for the optional selection systems which have been applied to the optimal selection for fine chemical experiment conditions is discussed.The application results are very satisfactory.

关 键 词：神经网络 实验方案 优选系统 化工实验 BP网络 

分 类 号：TQ016[化学工程] TP18[自动化与计算机技术—控制理论与控制工程]