Managing Training Examples for Fast Learning of Classifiers Ranks.

Boris Omelaenko, Vagan Terziyan, Seppo Puuronen: Managing Training Examples for Fast Learning of Classifiers Ranks CSIT 1999 : 139-148

Abstract

Paper deals with the problem of learning ranks of classifiers in ensembles. The problem of ordering of objects to classify is discussed. Two marginal approaches for learning, batch and incremental, with corresponding ordering strategies are analyzed. Presented algorithm lays between marginal methods, and it orders training examples by the deviation of classifiers opinions to match restrictions on learning time, cost and quality. Few aspects of this algorithm are experimentally investigated: classifiers ranks after learning, learning quality, ensemble accuracy and dependence between rank recalculation budget and ensemble accuracy. It was found, that descending order of examples provides fast rank learning with the best learning quality.

Copyright © 1999 by the Institute for Contemporary Education "JurInfoR-MSU". Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the CSIT copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the Institute for Contemporary Education JMSUICE. To copy otherwise, or to republish, requires a fee and/or special permission from the JMSUICE.

Printed Edition

Ch. Freytag and V. Wolfengagen (Eds.): CSIT'99, Proceedings of 1st International Workshop on Computer Science and Information Technologies, January 18-22, 1999, Moscow, Russia. MEPhI Publishing 1999, ISBN 5-7262-0263-5

Electronic Edition