Semi-supervised learning is used to study how to improve performance in the presence of both examples and instances,and it has become a hot area of machine learning field.In view of the theoretical significance and practical value of semi-supervised learning,semi-supervised learning methods were reviewed in this paper systematically.Firstly,some concepts about semi-supervised learning were summarized,including definition of semi-supervised learning,development of research,assumptions relied on semi-supervised learning methods and classification of semi-supervised learning. Secondly, semi-supervised learning methods were detailed from four aspects,including classification, regression, clustering, and dimension reduction. Thirdly, theoretical analysis on semi-supervised learning was studied,and error bounds and sample complexity were given.Finally,the future research on semi-supervised learning was discussed.
NETL
NSTL
万方数据
维普
