Back to Search Start Over

Rounding Methods for Discrete Linear Classification (Extended Version)

Authors :
Chevaleyre, Yann
Koriche, Frederic
Zucker, Jean-Daniel
Laboratoire d'Informatique de Paris-Nord (LIPN)
Université Sorbonne Paris Cité (USPC)-Institut Galilée-Université Paris 13 (UP13)-Centre National de la Recherche Scientifique (CNRS)
Centre de Recherche en Informatique de Lens (CRIL)
Université d'Artois (UA)-Centre National de la Recherche Scientifique (CNRS)
Institut de Recherche pour le Développement (IRD)
Chevaleyre, Yann
Université Paris 13 (UP13)-Institut Galilée-Université Sorbonne Paris Cité (USPC)-Centre National de la Recherche Scientifique (CNRS)
Centre National de la Recherche Scientifique (CNRS)-Université d'Artois (UA)
Publication Year :
2013
Publisher :
HAL CCSD, 2013.

Abstract

Learning discrete linear classifiers is known as a difficult challenge. In this paper, this learning task is cast as combinatorial optimization problem: given a training sample formed by positive and negative feature vectors in the Euclidean space, the goal is to find a discrete linear function that minimizes the cumulative hinge loss of the sample. Since this problem is NP-hard, we examine two simple rounding algorithms that discretize the fractional solution of the problem. Generalization bounds are derived for several classes of binary-weighted linear functions, by analyzing the Rademacher complexity of these classes and by establishing approximation bounds for our rounding algorithms. Our methods are evaluated on both synthetic and real-world data.

Details

Language :
English
Database :
OpenAIRE
Accession number :
edsair.dedup.wf.001..39a26c9070dd9a96d8b29fe5bf4db3c9