一种基于领域优化的集成交叉覆盖神经网络  

A cross cover neural network ensemble based on coverage areas optimization algorithm

作　　者：孙冰[1] 李晓丽[1] 

机构地区：[1]南京工业大学电子与信息工程学院,江苏南京210009

出　　处：《合肥工业大学学报（自然科学版）》2014年第2期238-242,共5页Journal of Hefei University of Technology：Natural Science

基　　金：高等学校博士学科点专项科研基金资助项目(20113221120003);江苏省高校自然科学基金资助项目(09KJB520006;11KJD520007)

摘　　要：文章介绍了交叉覆盖算法的基本思想,并对该算法的样本误识原因进行了分析,给出单个神经元的误识区域;在此基础上,提出了一种基于领域优化的集成交叉覆盖神经网络,在子网络的生成过程中,监视覆盖领域的生成情况,加入覆盖领域优化算法CAOA,避免集成后系统中出现覆盖领域重合、内切和外切的情况,以减小集成后系统的样本误识区域,提高系统的泛化性能。结果表明:采用基于CAOA的集成交叉覆盖神经网络可以减少拒识样本数,提高识别精度,对模式识别问题具有较强的实用价值。T he basic idea of the cross cover algorithm （CCA ） is introduced ,the reason of CCA sample misrecognition is analyzed and the misrecognition area of single neuron is given .On this basis ,a cross cover neural network ensemble based on coverage areas optimization algorithm （CAOA ）is put for-ward .In the process of creating subnetwork ,CAOA should be added to avoid coinciding ,inscribing and circumscribing of coverage areas .CAOA can decrease the sample misrecognition area of neural network ensemble ,and improve the recognition rate and generalization performance .The simulated experiment results show that the cross cover neural network ensemble based on CAOA can reduce the number of rejection samples and improve the identification accuracy ,and it is valuable for pattern rec-ognition problem .

关 键 词：交叉覆盖 神经网络集成 领域优化 泛化性能 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]