A Similarity-based Framework for Classification Task

Similarity-based method gives rise to a new class of methods for multi-label learning and also achieves promising performance. In this paper, we generalize method, resulting in framework classification task. Specifically, unite similarity-based generalized linear models achieve the best both worlds. This allows us capture interdependencies between classes prevent from impairing performance noisy classes. Each learned parameter model can reveal contribution one another, providing interpretability some extent. Experiment results show effectiveness proposed approach on multi-class data sets.