基于BP神经网络的改进算法研究

Study of improving algorithms based on the BP neural network

文章介绍了传统的BP算法,分析了它的不足之处,详细研究了BP算法的各种改进方法和措施,并通过实例,运用Matlab语言对部分改进算法进行了比较研究。实验结果表明,Levenberg-Marquardt自适应调整算法性能优于其它结合算法,可为其它工程提供参考。

<Abstrcat>The classic back-propagation(BP) learning algorithm of neural network and its disadvantages are introduced briefly. Some measures for improving the BP algorithm are studied by computer simulation using Matlab, and the experiment result show that Levenberg-Marquardt self-adaptation optimization algorithm is better than the others integrated.