摘要
神经网络分类器已被广泛应用在自动模式识别中。降低输入数据特征维数对其结构的简化和性能的提高至关重要。简单遗传算法早熟收敛和局部搜索能力弱的缺陷,使它在特征选择中的效果不理想。提出基于进化群体中值信息的动态自适应遗传算法。仿真结果表明,该算法优选特征子集速度快,解的质量稳定,神经网络分类器的识别准确率有显著提高。
Neural network classifier has been widely used in automatic recognition. It＇s performance and structure are dependent on the dimension of the input data. Genetic algorithm is a kind of global optimization algorithm and is appropriate to be used in feature selection. Two drawbacks of the Simple Genetic Algorithm （SGA）, i.e. premature convergence and poor local searching ab/lity, hinder SGA from wide application. Ttds paper proposed a novel genetic algorithm in which the median of the population are used to control the probability of the crossover and the mutation. Simulation results show that our method can get the optimal subset quickly and the performance of the neural network classifier could be improved markedly.
出处
《计算机应用》
CSCD
北大核心
2007年第9期2270-2272,共3页
journal of Computer Applications
关键词
遗传算法
特征选择
中值
神经网络分类器
genetic algorithm
feature selection
median
neural network classifier
