The Community for Technology Leaders
Green Image
Issue No. 04 - April (2012 vol. 34)
ISSN: 0162-8828
pp: 825-832
Chunhua Shen , The University of Adelaide, Adelaide
Hanzi Wang , Xiamen University, Xiamen
Peng Wang , Beihang University, Beijing
Fumin Shen , Nanjing University of Science and Technology, Nanjing
It has been shown that the Universum data, which do not belong to either class of the classification problem of interest, may contain useful prior domain knowledge for training a classifier [1], [2]. In this work, we design a novel boosting algorithm that takes advantage of the available Universum data, hence the name {\cal U}Boost. {\cal U}Boost is a boosting implementation of Vapnik's alternative capacity concept to the large margin approach. In addition to the standard regularization term, {\cal U}Boost also controls the learned model's capacity by maximizing the number of observed contradictions. Our experiments demonstrate that {\cal U}Boost can deliver improved classification accuracy over standard boosting algorithms that use labeled data alone.
Universum, kernel methods, boosting, column generation, convex optimization.
Chunhua Shen, Hanzi Wang, Peng Wang, Fumin Shen, "{\cal U}Boost: Boosting with the Universum", IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 34, no. , pp. 825-832, April 2012, doi:10.1109/TPAMI.2011.240
81 ms
(Ver 3.3 (11022016))