Workshop Invited Speaker

Maximum Margin Semi-supervised Learning with Irrelevant Data


Irwin King

Department of Computer Science and Engineering
The Chinese University of Hong Kong

  • [November 6-8, 2009, misc | www]
    Irwin King, {Maximum Margin Semi-supervised Learning With Irrelevant Data, Invited Talk at The 7th Chinese Workshop on Machine Learning and Applications (MLA'09), Nanjing University, Nanjing, China}, November 6-8, 2009.


Previous semi-supervised learning techniques usually assume the unlabeled data are relevant to the target task. In this talk, we address a different scenario, where the labeled and the unlabeled data may be a mixture of relevant or irrelevant data to the target binary classification task. Without specifying the relatedness in the unlabeled data, we develop a novel maximum margin classifier, named the tri-class support vector machine (3C-SVM), to seek an inductive rule that can separate these data into three categories: -1, +1, or 0. A key to fulfill the 3C-SVM is that we introduce a new min loss function. This loss function achieves the maximum entropy principle and therefore can distinguish the unlabeled data into relevant and irrelevant data. The 3C-SVM can then generalize standard SVMs, Semi-supervised SVMs, and SVMs learned from the universum as its special cases. For implementation, the 3C-SVM just needs to solve several quadratic programming problems, which there exist some highly scalable algorithms. Results on both synthetic datasets and two benchmark handwritten digit datasets are reported to demonstrate the advantage of the 3C-SVM algorithm.

Brief Profile

Dr. King's research interests include machine learning, web intelligence & social computing, and multimedia processing. In these research areas, he has over 160 technical publications in journals (JMLR, ACM TOIS, IEEE TNN, Neurocomputing, NN, IEEE BME, PR, IEEE SMC, JAMC, JASIST, IJPRAI, DSS, etc.) and conferences (NIPS, IJCAI, CIKM, SIGIR, KDD, PAKDD, ICDM, WWW, WI/IAT, WCCI, IJCNN, ICONIP, ICDAR, etc.). In addition, he has contributed over 20 book chapters and edited volumes. Moreover, Dr. King has over 30 research and applied grants. One notable system he has developed is the VeriGuide system, which detects similar sentences and performs readability analysis of text-based documents in both English and in Chinese to promote academic integrity and honesty.

Dr. King is an Associate Editor of the IEEE Transactions on Neural Networks (TNN) and IEEE Computational Intelligence Magazine (CIM). He is a member of the Editorial Board of the Open Information Systems Journal, Journal of Nonlinear Analysis and Applied Mathematics, and Neural Information Processing–Letters and Reviews Journal (NIP-LR). He has also served as Special Issue Guest Editor for Neurocomputing, International Journal of Intelligent Computing and Cybernetics (IJICC), Journal of Intelligent Information Systems (JIIS), and International Journal of Computational Intelligent Research (IJCIR). He is a senior member of IEEE and a member of ACM, International Neural Network Society (INNS), and Asian Pacific Neural Network Assembly (APNNA). Currently, he is serving the Neural Network Technical Committee (NNTC) and the Data Mining Technical Committee under the IEEE Computational Intelligence Society (formerly the IEEE Neural Network Society). He is also a Vice-President and Governing Board Member of the Asian Pacific Neural Network Assembly (APNNA).

Dr. King joined the Chinese University of Hong Kong in 1993. He received his B.Sc. degree in Engineering and Applied Science from California Institute of Technology, Pasadena and his M.Sc. and Ph.D. degree in Computer Science from the University of Southern California, Los Angeles.

confs/mla2009.txt · Last modified: 2009/11/11 01:00 (external edit)     Back to top