摘要
主要阐述了一种新的遗传算法-自适应递阶遗传算法的基本原理,解决了长期以来无法同时对神经网络拓扑结构和神经网络的权值和阈值进行优化的问题,设计了一个基于自适应递阶遗传算法的BP神经网络学习算法,给出了具体的程序设计,并且利用MATLAB平台进行仿真计算.实验结果表明,该算法比一般遗传算法具有明显的优越性,可以避免神经网络陷入局部最优,迅速优化网络的拓扑结构,提高了网络的学习性能,具有一定的实用性。
The paper proposes the basic principle of a new genetic algorithm adaptive hierarchical genetic algorithm (AHGA), solves the problem of optimizing the structure and the weights of neural network simultaneously, and gives the specific program design, The result of experiment based on MATLAB indicates that AHGA algorithm has superiority over simple genetic algorithm (SGA), It can avoid the disadvantages of BP algorithm ,optimize the structure of network quickly, improve the learning performance of network.
出处
《计算机仿真》
CSCD
2007年第8期159-162,共4页
Computer Simulation
基金
国家十五攻关子课题(2004BA204B08-03)
关键词
神经网络
自适应递阶遗传算法
优化
Neural network
Adaptive hierarchical genetic algorithm
Optimization