A New Framework for Multiclass Classification Using Multiview Assisted Adaptive Boosting

摘要: Multiview representation of data is common in disciplines such as computer vision, bio-informatics, etc. Traditional fusion methods train independent classifiers on each view and finally conglomerate them using weighted summation. Such approaches are void from inter-view communications thus do not guarantee to yield the best possible ensemble classifier given sample-view space. This paper proposes a new algorithm for multiclass classification multi-view assisted supervised learning (MA-AdaBoost). MA-AdaBoost uses adaptive boosting initially training baseline view. After round, share their performances. Based this communication, weight an example ascertained by its difficulties across all views. Two versions proposed based nature final output classifiers. Finally, decisions agglomerated novel reward assignment. The then presents comparisons benchmark UCI datasets eye samples collected FERET database. Kappa-error diversity diagrams also studied. In majority instances, outperforms traditional AdaBoost, variants recent works collaborative with respect convergence rate set generalization errors. error-diversity results encouraging.