Local regularization for multiclass classification facing significant intraclass variations

Lior Wolf*, Yoni Donner

*Corresponding author for this work

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

Abstract

We propose a new local learning scheme that is based on the principle of decisiveness: the learned classifier is expected to exhibit large variability in the direction of the test example. We show how this principle leads to optimization functions in which the regularization term is modified, rather than the empirical loss term as in most local learning schemes. We combine this local learning method with a Canonical Correlation Analysis based classification method, which is shown to be similar to multiclass LDA. Finally, we show that the classification function can be computed efficiently by reusing the results of previous computations. In a variety of experiments on new and existing data sets, we demonstrate the effectiveness of the CCA based classification method compared to SVM and Nearest Neighbor classifiers, and show that the newly proposed local learning method improves it even further, and outperforms conventional local learning schemes.

Original languageEnglish
Title of host publicationComputer Vision - ECCV 2008 - 10th European Conference on Computer Vision, Proceedings
PublisherSpringer Verlag
Pages748-759
Number of pages12
EditionPART 4
ISBN (Print)3540886923, 9783540886921
DOIs
StatePublished - 2008
Event10th European Conference on Computer Vision, ECCV 2008 - Marseille, France
Duration: 12 Oct 200818 Oct 2008

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 4
Volume5305 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference10th European Conference on Computer Vision, ECCV 2008
Country/TerritoryFrance
CityMarseille
Period12/10/0818/10/08

Fingerprint

Dive into the research topics of 'Local regularization for multiclass classification facing significant intraclass variations'. Together they form a unique fingerprint.

Cite this