A Semi-Supervised Framework for Feature Mapping and Multiclass Classification

Publisher:
SIAM
Publication Type:
Conference Proceeding
Citation:
Proceedings of SIAM International Conference on Data Mining, 2009
Issue Date:
2009-01-01
Full metadata record
Files in This Item:
Filename Description Size
2013006913OK.pdf Published version1.35 MB
Adobe PDF
We propose a semi-supervised framework incorporat- ing feature mapping with multiclass classi¿cation. By learning multiple classi¿cation tasks simultaneously, this framework can learn the latent feature space e¿ec- tively for both labeled and unlabeled data. The knowl- edge in the transformed space can be transferred not only between the labeled and unlabeled data, but also across multiple classes, so as to improve the classi¿ca- tion performance given a small amount of labeled data. We show that this problem is equivalent to a sequen- tial convex optimization problem by applying constraint concave-convex procedure (CCCP). E¿cient algorithm with theoretical guarantee is proposed and computa- tional issue is investigated. Extensive experiments have been conducted to demonstrate the e¿ectiveness of our proposed framework.
Please use this identifier to cite or link to this item: