Consistency Based Partial Label Learning Algorithm
TANG Caizhi, ZHANG Minling
School of Computer Science and Engineering, Southeast University, Nanjing 210096 Key Laboratory of Computer Network and Information Integration, Ministry of Education,Southeast University, Nanjing 210096
Abstract：An essential strategy to solve the partial label problem is disambiguation. In most existing strategies, instances are individually disambiguated without the consideration of the relationships among instances. In this paper, a consistency based partial label learning (COPAL) algorithm is proposed assumpting that labels associated with similar instances are likely to be similar. Based on the above assumption, the labeling information of the instance itself and its neighboring instances are simultaneously utilized for disambiguation. Experiments on both artificial datasets and realworld datasets show the good generalization ability of COPAL.
[1] LUO J, ORABONA F. Learning from Candidate Labeling Sets [EB/OL]. [2016-02-10]. http://www.luojie.me/publications/files/lj_fo-nips10.pdf. [2] CHAPELLE O, SCH LKOPF B, ZIEN A. Semi-supervised Lear-ning. Cambridge, USA: MIT Press, 2006. [3] ZHANG M L, ZHOU Z H. A Review on Multi-label Learning Algorithms. IEEE Trans on Knowledge and Data Engineering, 2014, 26(8): 1819-1837. [4] AMORES J. Multiple Instance Classification: Review, Taxonomy and Comparative Study. Artificial Intelligence, 2013, 201: 81-105. [5] ZHANG M L. Disambiguation-Free Partial Label Learning [EB/OL]. [2016-02-10]. http://epubs.siam.org/doi/pdf/10.1137/1.9781611973440.5. [6] JIN R, GHAHRAMANI Z. Learning with Multiple Labels // THRUNS S, OBERMAYER K, eds. Advances in Neural Information Processing Systems 15. Cambridge, USA: MIT Press, 2002: 897-904. [7] LIU L P, DIETTERICH T G. A Conditional Multinomial Mixture Model for Superset Label Learning // PEREIRA F, BURGES C J C, BOTTOU L, et al., eds. Advances in Neural Information Processing Systems 25. Cambridge, USA: MIT Press, 2012: 557-565. [8] NGUYEN N, CARUANA R. Classification with Partial Labels // Proc of the 14th ACM SIGKDD International Conference on Know-ledge Discovery and Data Mining. New York, USA: ACM, 2008: 551-559. [9] YU F, ZHANG M L. Maximum Margin Partial Label Learning // Proc of the 7th Asian Conference on Machine Learning. Cambridge, USA: MIT Press, 2015: 96-111. [10] HLLERMEIER E, BERINGER J. Learning from Ambiguously Labeled Examples // Proc of the 6th International Conference on Advances in Intelligent Data Analysis. Berlin, Germany: Springer-Verlag, 2005: 168-179. [11] COUR T, SAPP B, TASKAR B. Learning from Partial Labels. Journal of Machine Learning Research, 2011, 12: 1501-1536. [12] ZHANG M L, YU F. Solving the Partial Label Learning Problem: An Instance-Based Approach // Proc of the 24th International Joint Conference on Artificial Intelligence. Palo Alto, USA: AAAI Press, 2015: 4048-4054. [13] MALLAPRAGADA P K, JIN R, JAIN A K, et al. Semiboost: Boosting for Semi-supervised Learning. IEEE Trans on Pattern Ana-lysis and Machine Intelligence, 2009, 31(11): 2000-2014. [14] LIU D C, NOCEDAL J. On the Limited Memory BFGS Method for Large Scale Optimization. Mathematical Programming, 1989, 45(3): 503-528. [15] BRIGGS F, FERN X Z, RAICH R. Rank-Loss Support Instance Machines for MIML Instance Annotation // Proc of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2012: 534-542. [16] HAN H, OTTO C, LIU X M, et al. Demographic Estimation from Face Images: Human vs. Machine Performance. IEEE Trans on Pa-ttern Analysis and Machine Intelligence, 2015, 37(6): 1148-1161. [17] LANITIS A, TAYLOR C J, COOTES T F. Toward Automatic Simulation of Aging Effects on Face Images. IEEE Trans on Pattern Ana-lysis and Machine Intelligence, 2002, 24(4): 442-455.
2016, Vol. 29 
