In applications where categorical labels follow a natural hierarchy, classification methods that exploit the label structure often outperform those that do not. Un-fortunately, the majority of classification datasets do not come pre-equipped with a hierarchical structure and classical flat classifiers must be employed. In this paper, we investigate a class of methods that induce a hierarchy that can similarly improve classification performance over flat classifiers. The class of methods follows the structure of first clustering the conditional distributions and subsequently using a hierarchical classifier with the induced hierarchy. We demonstrate the effectiveness of the class of methods both for discovering a latent hierarchy and for improving accuracy in principled simulation settings and three real data applications.
Classification Fields: Arbitrarily Fine Recursive Hierarchical Clustering From Few Examples
Department of Mathematics and Statistics, McMaster University, Canada（加拿大麦 master 大学数学与统计学系）；Vector Institute, Canada（加拿大向量研究所）；University of Cambridge, United Kingdom（英国剑桥大学）
Ranking hierarchical multi-label classification results with mLPRs
Institute of Statistics and Data Science, National Taiwan University, Taipei, Taiwan.（统计与数据科学研究所，国立台湾大学，台北，台湾）；Department of Statistics University of California Berkeley, CA, USA.（统计学系，加州大学伯克利分校，加州，美国）
Multi-task Learning of Pairwise Sequence Classification Tasks Over Disparate Label Spaces
Department of Computer Science, University of Copenhagen（哥本哈根大学计算机科学系）；Insight Research Centre, National University of Ireland, Galway（爱尔兰国立大学戈尔韦分校Insight研究中心）；Aylien Ltd., Dublin（Aylien有限公司）
