中国计量大学学报2017，Vol.28Issue(4)：409-417,9.DOI:10.3969/j.issn.2096-2835.2017.04.001
跨领域分布适配超限学习机及其在域自适应问题的应用
Domain distribution matching extreme learning machines for domain adaptation problems
摘要
Abstract
Domain adaptation is an effective solution for domain shift problems,where the training data(source domain)and testing data(target domain)are drawn from the different but related distributions.In this paper, we propose a domain distribution matching extreme learning machine(DDM-ELM)method for domain adaptation tasks.DDM-ELM aims to fully leverage rich labeled source data and unlabeled target data to build an accurate classifier for the target domain based on ELM frameworks.Specifically,DDM-ELM simultaneously minimizes the classification error of the source data,effectively reduces the distribution distance between the source and target domains by minimizing the projected maximum mean discrepancy,and explores the structure property of the target domain by using manifold regularization.This leads DDM-ELM to be more adaptive to the target domain and maintains the merits of ELM.Extensive experiments verify that DDM-ELM are competitive with several state-of-the-art domain adaptation methods in terms of the classification accuracy and efficiency.关键词
模式识别/数据挖掘/域自适应/超限学习机Key words
pattern recognition/data mining/domain adaptation/extreme learning machine分类
信息技术与安全科学引用本文复制引用
宋士吉,李爽..跨领域分布适配超限学习机及其在域自适应问题的应用[J].中国计量大学学报,2017,28(4):409-417,9.基金项目
教育部高等学校博士点基金资助项目(No.20130002130010). （No.20130002130010）
