Skip to Main content Skip to Navigation
Book sections

Maximum-Margin Framework for Training Data Synchronization in Large-Scale Hierarchical Classification

Abstract : In the context of supervised learning, the training data for large-scale hierarchical classification consist of (i) a set of input-output pairs, and (ii) a hierarchy structure defining parent-child relation among class labels. It is often the case that the hierarchy structure given a-priori is not optimal for achieving high classification accuracy. This is especially true for web-taxonomies such as Yahoo! directory which consist of tens of thousand of classes. Furthermore, an important goal of hierarchy design is to render better navigability and browsing. In this work, we propose a maximum-margin framework for automatically adapting the given hierarchy by using the set of input-output pairs to yield a new hierarchy. The proposed method is not only theoretically justified but also provides a more principled approach for hierarchy flattening techniques proposed earlier, which are ad-hoc and empirical in nature. The empirical results on publicly available large-scale datasets demonstrate that classification with new hierarchy leads to better or comparable generalization performance than the hierarchy flattening techniques.
Complete list of metadatas

Cited literature [10 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-00943206
Contributor : Rohit Babbar <>
Submitted on : Friday, February 7, 2014 - 11:03:06 AM
Last modification on : Monday, April 20, 2020 - 11:24:01 AM
Document(s) archivé(s) le : Monday, May 12, 2014 - 11:55:42 AM

File

372.pdf
Files produced by the author(s)

Identifiers

Collections

CNRS | LIG | UGA

Citation

Rohit Babbar, Ioannis Partalas, Eric Gaussier, Massih-Reza Amini. Maximum-Margin Framework for Training Data Synchronization in Large-Scale Hierarchical Classification. Lee, Minho and Hirose, Akira and Hou, Zeng-Guang and Kil, RheeMan. Neural Information Processing, Springer Berlin Heidelberg, pp.336-343, 2013, ⟨10.1007/978-3-642-42054-2_42⟩. ⟨hal-00943206⟩

Share

Metrics

Record views

279

Files downloads

497