Set analysis of coincident errors and its applications for combining classifiers.

Ruta, D. and Gabrys, B., 2003. Set analysis of coincident errors and its applications for combining classifiers. In: Chen, D. and Cheng, X., eds. Pattern Recognition and String Matching. Dordrecht; Boston: Kluwer Academic, pp. 647-672.

Full text available as:

[img]
Preview
PDF
393kB

Abstract

Although addressed in many papers, classifier dependency is still not well defined. Continuously being described by variety of statistical models from conditional probability to diversity measures, dependency among classifier out-puts was recently shown to have a crucial impact on the performance of multi-ple classifier system. However, individual classifier performances still represent competitive and simple information clearly related to the performance of the combined system. In this work we show that all the measures related to classi-fier outputs can be reformulated to represent just different forms of the same in-formation of error coincidences. Applying set analysis for the representation and description of error coincidences we define collection of classifier sets de-composed into two complementary types of coincidence levels. Furthermore we illustrate a high flexibility of using the coincidence levels, which supported be a simple algebra cover many established dependency measures including combin-ing error in case of majority voting. Moreover we show that in the sets-collection representation of error coincidences a specific inclusion relation re-sults in a quicker and more effective handling of dependency information under different complexity conditions. In the experimental section we examine rela-tions of the introduced error coincidence levels with majority voting combiner using real datasets and classifiers and indicate further potential applications of the presented concepts.

Item Type:Book Section
ISBN:9781402009532
Series Name:Combinatorial Optimization
Volume:13
Number of Pages:772
Series Name:Combinatorial Optimization
Additional Information:This invited peer-reviewed book chapter is a result of our analysis of usefulness of various diversity measures used in the context of multiple classifier systems and presents a completely new view of coincidental errors based on set analysis. Although addressed in many papers, classifier dependency is still not well defined. Continuously being described by variety of statistical models from conditional probability to diversity measures, dependency among classifier outputs was shown to have a crucial impact on the performance of multiple classifier system. In this work we showed that all the measures related to classifier outputs can be reformulated to represent just different forms of the same information of error coincidences. Applying set analysis for the representation and description of error coincidences we defined a collection of classifier sets decomposed into two complementary types of coincidence levels. We also showed that in the sets collection representation of error coincidences a specific inclusion relation results in a quicker and more effective handling of dependency information under different complexity conditions.
Subjects:Generalities > Computer Science and Informatics > Artificial Intelligence
Generalities > Computer Science and Informatics
Group:School of Design, Engineering & Computing > Smart Technology Research Centre
ID Code:1086
Deposited By:INVALID USER
Deposited On:17 Dec 2007
Last Modified:07 Mar 2013 14:36

Document Downloads

More statistics for this item...
Repository Staff Only -
BU Staff Only -
Help Guide - Editing Your Items in BURO