Using the Geometrical Distribution of Prototypes for Training Set Condensing

Abstract

Abstract. In this paper, some new approaches to training set size reduction are presented. These schemes basically consist of defining a small number of prototypes that represent all the original instances. Although the ultimate aim of the algorithms proposed here is to obtain a strongly reduced training set, the performance is empirically evaluated over nine real datasets by comparing the reduction rate and the classification accuracy with those of other condensing techniques.

Trends in Nearest Feature Classification for Face Recognition—Achievements and Perspectives

“…Condensing algorithms try to select a small subset of prototypes while preserving classification performance as good as possible. Condensing may involve just a pure selection of prototypes (Hart, 1968;Tomek, 1976;Toussaint et al, 1985;Dasarathy, 1990;Dasarathy, 1994) or include a modification of them (Chang, 1974;Chen & Józwik, 1996;Ainslie & Sánchez, 2002;Lozano et al, 2004a;Lozano et al, 2004b).…”

Experimental study on prototype optimisation algorithms for prototype-based classification in vector spaces

“…The Reconsistent algorithm is an important modification of the MaxNCN algorithm towards obtaining a consistent condensed set [14]. The primary idea is that the consistency of a subset with respect to the TS should lead to a better classification.…”

“…Two main groups of condensing techniques can be distinguished. These are the selective schemes, which merely select a subset of the original training objects [2,[6][7][8][9] and the adaptive schemes which modify them [10][11][12][13][14][15].…”

“…These are MaxNCN [14,23], Reconsistent [14], LVQ [24] and MixtGauss [15]. Section 3 briefly describes the framework of the prototype-based classification methods used for the evaluation of the derived condensed sets.…”