Hierarchical Maximum Margin Learning for Multi-Class Classification

摘要

Due to myriads of classes, designing accurate and efficient classifiers becomes very chal lenging for multi-class classification. Recent research has shown that class structure learn ing can greatly facilitate multi-class learning. In this paper, we propose a novel method to learn the class structure for multi-class classification problems. The class structure is assumed to be a binary hierarchical tree. To learn such a tree, we propose a max imum separating margin method to deter mine the child nodes of any internal node. The proposed method ensures that two class groups represented by any two sibling nodes are most separable. In the experiments, we evaluate the accuracy and efficiency of the proposed method over other multi-class clas sification methods on real world large-scale problems. The results show that the pro posed method outperforms benchmark meth ods in terms of accuracy for most datasets and performs comparably with other class structure learning methods in terms of effi ciency for all datasets.