Efficient max-margin metric learning

Caiming Xiong, David Johnson, Jason Corso

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

Efficient learning of an appropriate distance metric is an increasingly important problem in machine learning. However, current methods are limited by scalability issues or are unsuited to use with general similarity/dissimilarity constraints. In this paper, we propose an efficient metric learning method based on the max-margin framework with pairwise constraints that has strong generalization gu arantees. First, we reformulate the max-margin metric learning problem as a structured support vector machine which we can optimize in linear time via a cutting-plane method. Second, we propose a kernelized extension to the method, with a linear-time-computable approximation based on a matching pursuit algorithm. We find our method to be comparable to or better than state of the art metric learning techniques at a number of machine learning and computer vision classification tasks.

Original languageEnglish
Title of host publicationProceedings of the IADIS International Conference Intelligent Systems and Agents 2012, ISA 2012, IADIS European Conference on Data Mining 2012, ECDM 2012
Pages115-123
Number of pages9
StatePublished - 2012
EventIADIS International Conference Intelligent Systems and Agents 2012, ISA 2012, IADIS European Conference on Data Mining 2012, ECDM 2012, Part of the IADIS Multi Conference on Computer Science and Information Systems 2012, MCCSIS 2012 - Lisbon, Portugal
Duration: 21 Jul 201223 Jul 2012

Publication series

NameProceedings of the IADIS International Conference Intelligent Systems and Agents 2012, ISA 2012, IADIS European Conference on Data Mining 2012, ECDM 2012

Conference

ConferenceIADIS International Conference Intelligent Systems and Agents 2012, ISA 2012, IADIS European Conference on Data Mining 2012, ECDM 2012, Part of the IADIS Multi Conference on Computer Science and Information Systems 2012, MCCSIS 2012
Country/TerritoryPortugal
CityLisbon
Period21/07/1223/07/12

Keywords

  • Cutting-Plane
  • Matching Pursuit
  • Metric Learning
  • Structured SVM

Fingerprint

Dive into the research topics of 'Efficient max-margin metric learning'. Together they form a unique fingerprint.

Cite this