BCN Perceptual Computing Lab Repository

Compact Evolutive Design of Error-Correcting Output Codes

Bautista, Miguel A, and Baró, Xavier and Pujol, Oriol and Radeva, Petia and Vitrià, Jordi and Escalera, Sergio (2010) Compact Evolutive Design of Error-Correcting Output Codes. Proceedings of the Supervised and Unsupervised Methods and their Applications (SUEMA), European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases .

[img]PDF (Paper) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
763Kb
[img]PDF (Slides) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
926Kb

Abstract

The classi�cation of large number of object categories is a challenging trend in the Machine Learning �eld. In literature, this is often addressed using an ensemble of classi�ers. In this scope, the Error- Correcting Output Codes framework has demonstrated to be a powerful tool for the combination of classi�ers. However, most of the state-of-the- art ECOC approaches use a linear or exponential number of classi�ers, making the discrimination of a large number of classes unfeasible. In this paper, we explore and propose a minimal design of ECOC in terms of the number of classi�ers. Evolutionary computation is used for tuning the parameters of the classi�ers and looking for the best Minimal ECOC code con�guration. The results over several public UCI data sets and a challenging multi-class Computer Vision problem show that the proposed methodology obtains comparable and even better results than state-of- the-art ECOC methodologies with far less number of dichotomizers.

Item Type:Article
Subjects:Q Science > QA Mathematics > QA75 Electronic computers. Computer science
ID Code:49
Deposited By:Xavier Baró
Deposited On:02 Dec 2010 12:44
Last Modified:02 Dec 2010 12:44

Repository Staff Only: item control page