摘要
介绍神经网络集成的基本概念及其算法理论,设计了改进的自适应交叉概率和变异概率,提出基于种群适应度集中程度的改进遗传算法,利用该算法优化加权平均集成网络的权,从而构建一种高效的神经网络集成模型。该模型用于解决分类问题.表现出比传统神经网络集成模型更好的性能。
It introduced the basic concept and its algorithm theory of neural network ensembles, designed the improved adaptive crossover probability and mutation probability, proposed an improved genetic algorithm, according to the concentrating degree of fitness of the populations. It used the algorithm to optimize neural network ensembles weight, in order to build an efficient model of neural network ensembles. This model is compared with standard ensembles in solving several problems of classification with excellent performance.
出处
《微计算机信息》
2010年第33期206-207,234,共3页
Control & Automation
关键词
遗传算法
神经网络集成
自适应交叉概率
自适应变异概率
genetic algorithm
neural network ensembles
adaptive crossover probability
adaptive mutation probability
