Copyright 2003 ACM. Automatic classification by machines is one of the basic tasks required in any pattern recognition and human computer interaction applications. In this paper, we discuss training probabilistic classifiers with labeled and unlabeled data. We provide an analysis which shows under what conditions unlabeled data can be used in learning to improve classification performance. We discuss the implications of this analysis to a specific type of probabilistic classifiers, Bayesian networks, and propose a structure learning algorithm that can utilize unlabeled data to improve classification. Finally, we show how the resulting algorithms are successfully employed in a facial expression recognition application.
|Original language||American English|
|Number of pages||6|
|State||Published - 7 Nov 2003|
|Event||Proceedings of the 5th ACM SIGMM International Workshop on Multimedia Information Retrieval, MIR 2003 - |
Duration: 7 Nov 2003 → …
|Conference||Proceedings of the 5th ACM SIGMM International Workshop on Multimedia Information Retrieval, MIR 2003|
|Period||7/11/03 → …|