Majority Vote of Diverse Classifiers for Late Fusion

In the past few years, a lot of attention has been devoted to multimedia indexing by fusing multimodal informations. Two kinds of fusion schemes are generally considered: The early fusion and the late fusion. We focus on late classifier fusion, where one combines the scores of each modality at the decision level. To tackle this problem, we investigate a recent and elegant well-founded quadratic program named MinCq coming from the machine learning PAC-Bayesian theory. MinCq looks for the weighted combination, over a set of real-valued functions seen as voters, leading to the lowest misclassification rate, while maximizing the voters' diversity. We propose an extension of MinCq tailored to multimedia indexing. Our method is based on an order-preserving pairwise loss adapted to ranking that allows us to improve Mean Averaged Precision measure while taking into account the diversity of the voters that we want to fuse. We provide evidence that this method is naturally adapted to late fusion procedures and confirm the good behavior of our approach on the challenging PASCAL VOC'07 benchmark.

[1]  Tong Zhang,et al.  Statistical Analysis of Some Multi-Category Large Margin Classification Methods , 2004, J. Mach. Learn. Res..

[2]  Eyke Hüllermeier,et al.  Preference Learning , 2005, Künstliche Intell..

[3]  Eyke Hllermeier,et al.  Preference Learning , 2010 .

[4]  Mohan S. Kankanhalli,et al.  Multimodal fusion for multimedia analysis: a survey , 2010, Multimedia Systems.

[5]  Edward Y. Chang,et al.  Optimal multimodal fusion for multimedia data analysis , 2004, MULTIMEDIA '04.

[6]  Jiri Matas,et al.  On Combining Classifiers , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[7]  Chih-Jen Lin,et al.  LIBSVM: A library for support vector machines , 2011, TIST.

[8]  Shiliang Sun,et al.  A survey of multi-view machine learning , 2013, Neural Computing and Applications.

[9]  Luc Van Gool,et al.  The Pascal Visual Object Classes (VOC) Challenge , 2010, International Journal of Computer Vision.

[10]  Yoav Freund,et al.  Experiments with a New Boosting Algorithm , 1996, ICML.

[11]  ZissermanAndrew,et al.  The Pascal Visual Object Classes Challenge , 2015 .

[12]  Luc Van Gool,et al.  The 2005 PASCAL Visual Object Classes Challenge , 2005, MLCW.

[13]  Subhash C. Bagui,et al.  Combining Pattern Classifiers: Methods and Algorithms , 2005, Technometrics.

[14]  François Laviolette,et al.  From PAC-Bayes Bounds to Quadratic Programs for Majority Votes , 2011, ICML.

[15]  Giorgio Valentini,et al.  Ensemble methods : a review , 2012 .

[16]  László Györfi,et al.  A Probabilistic Theory of Pattern Recognition , 1996, Stochastic Modelling and Applied Probability.

[17]  Jianguo Zhang,et al.  The PASCAL Visual Object Classes Challenge , 2006 .

[18]  Bernard F. Buxton,et al.  Performance Degradation in Boosting , 2001, Multiple Classifier Systems.

[19]  David A. McAllester PAC-Bayesian model averaging , 1999, COLT '99.

[20]  Patrick Gallinari,et al.  Multiview semi-supervised ranking for automatic image annotation , 2013, MM '13.

[21]  Ashok N. Srivastava,et al.  Advances in Machine Learning and Data Mining for Astronomy , 2012 .

[22]  Filip Radlinski,et al.  A support vector method for optimizing average precision , 2007, SIGIR.

[23]  Ludmila I. Kuncheva,et al.  Combining Pattern Classifiers: Methods and Algorithms , 2004 .

[24]  François Laviolette,et al.  PAC-Bayes Bounds for the Risk of the Majority Vote and the Variance of the Gibbs Classifier , 2006, NIPS.

[25]  Thomas G. Dietterich Multiple Classifier Systems , 2000, Lecture Notes in Computer Science.

[26]  David H. Wolpert,et al.  Stacked generalization , 1992, Neural Networks.

[27]  Jian-Huang Lai,et al.  Linear Dependency Modeling for Classifier Fusion and Feature Combination , 2013, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[28]  John Dunnion,et al.  Applying Machine Learning Diversity Metrics to Data Fusion in Information Retrieval , 2011, ECIR.