Selecting Critical Patterns Based on Local Geometrical and Statistical Information

Yuhua Li, LP Maguire

Research output: Non-textual formWeb publication/site

Abstract

Pattern selection methods have been traditionally developed with a dependency on a specific classifier. In contrast this paper presents a method that selects critical patterns deemed to carry essential information applicable to train those types of classifiers which require spatial information of the training dataset. Critical patterns include those edge patterns that define the boundary and those border patterns that separate classes. The proposed method selects patterns from a new perspective, primarily based on their location in input space. It determines class edge patterns with the assistance of approximated tangent hyperplane of a class surface. It also identifies border patterns between classes using local probability. The proposed method is evaluated on benchmark problems using popular classifiers including multilayer perceptrons, radial basis functions, support vector machines and nearest neighbors. The proposed approach is also compared with four state-of-the-art approaches and it is shown to provide similar but more consistent accuracy from a reduced data set. Experimental results demonstrate that it selects patterns sufficient to represent class boundary and to preserve the decision surface.
LanguageEnglish
DOIs
Publication statusPublished - 2010

Fingerprint

Classifiers
Multilayer neural networks
Support vector machines

Cite this

@misc{217abb28ce8c40638ba1ca6152bac7d8,
title = "Selecting Critical Patterns Based on Local Geometrical and Statistical Information",
abstract = "Pattern selection methods have been traditionally developed with a dependency on a specific classifier. In contrast this paper presents a method that selects critical patterns deemed to carry essential information applicable to train those types of classifiers which require spatial information of the training dataset. Critical patterns include those edge patterns that define the boundary and those border patterns that separate classes. The proposed method selects patterns from a new perspective, primarily based on their location in input space. It determines class edge patterns with the assistance of approximated tangent hyperplane of a class surface. It also identifies border patterns between classes using local probability. The proposed method is evaluated on benchmark problems using popular classifiers including multilayer perceptrons, radial basis functions, support vector machines and nearest neighbors. The proposed approach is also compared with four state-of-the-art approaches and it is shown to provide similar but more consistent accuracy from a reduced data set. Experimental results demonstrate that it selects patterns sufficient to represent class boundary and to preserve the decision surface.",
author = "Yuhua Li and LP Maguire",
year = "2010",
doi = "10.1109/TPAMI.2010.18",
language = "English",

}

Selecting Critical Patterns Based on Local Geometrical and Statistical Information. Li, Yuhua (Author); Maguire, LP (Author). 2010.

Research output: Non-textual formWeb publication/site

TY - ADVS

T1 - Selecting Critical Patterns Based on Local Geometrical and Statistical Information

AU - Li, Yuhua

AU - Maguire, LP

PY - 2010

Y1 - 2010

N2 - Pattern selection methods have been traditionally developed with a dependency on a specific classifier. In contrast this paper presents a method that selects critical patterns deemed to carry essential information applicable to train those types of classifiers which require spatial information of the training dataset. Critical patterns include those edge patterns that define the boundary and those border patterns that separate classes. The proposed method selects patterns from a new perspective, primarily based on their location in input space. It determines class edge patterns with the assistance of approximated tangent hyperplane of a class surface. It also identifies border patterns between classes using local probability. The proposed method is evaluated on benchmark problems using popular classifiers including multilayer perceptrons, radial basis functions, support vector machines and nearest neighbors. The proposed approach is also compared with four state-of-the-art approaches and it is shown to provide similar but more consistent accuracy from a reduced data set. Experimental results demonstrate that it selects patterns sufficient to represent class boundary and to preserve the decision surface.

AB - Pattern selection methods have been traditionally developed with a dependency on a specific classifier. In contrast this paper presents a method that selects critical patterns deemed to carry essential information applicable to train those types of classifiers which require spatial information of the training dataset. Critical patterns include those edge patterns that define the boundary and those border patterns that separate classes. The proposed method selects patterns from a new perspective, primarily based on their location in input space. It determines class edge patterns with the assistance of approximated tangent hyperplane of a class surface. It also identifies border patterns between classes using local probability. The proposed method is evaluated on benchmark problems using popular classifiers including multilayer perceptrons, radial basis functions, support vector machines and nearest neighbors. The proposed approach is also compared with four state-of-the-art approaches and it is shown to provide similar but more consistent accuracy from a reduced data set. Experimental results demonstrate that it selects patterns sufficient to represent class boundary and to preserve the decision surface.

U2 - 10.1109/TPAMI.2010.18

DO - 10.1109/TPAMI.2010.18

M3 - Web publication/site

ER -