A class discriminality measure based on feature space partitioning

Authors:
AndréF. Kohn;Luis G. M. Nakano;Miguel Oliveira E Silva
Affiliations:
Laboratório de Engenharia Biomédica, Departamento de Engenharia Electronica, Escola Politécnica, Universidade de Sao Paulo, Cx. P. 61548, CEP 05424-970, Sao Paulo, S. P., Brazil;Laboratório de Sistemas Digitais, Departamento de Engenharia de Computa¢ão, Escola Politecnica, Universidade de Sao Paulo, Brazil;Departamento de Electrónica e Telecomunica¢ões, Universidade de Aveiro, Portugal
Venue:
Pattern Recognition
Year:
1996

Citing 3
Cited 3

Bayes Error Estimation Using Parzen and k-NN Procedures

IEEE Transactions on Pattern Analysis and Machine Intelligence
Feature Extraction Based on Decision Boundaries

IEEE Transactions on Pattern Analysis and Machine Intelligence
Optimization of k nearest neighbor density estimates

IEEE Transactions on Information Theory

Multiresolution Estimates of Classification Complexity

IEEE Transactions on Pattern Analysis and Machine Intelligence
Addressing the Problems of Bayesian Network Classification of Video Using High-Dimensional Features

IEEE Transactions on Knowledge and Data Engineering
Multiresolution estimates of classification complexity and multiple subspace classifiers for understanding and solving complex recognition tasks

SPPRA'06 Proceedings of the 24th IASTED international conference on Signal processing, pattern recognition, and applications

Quantified Score

Hi-index	0.01

Visualization

Abstract

This paper presents a new class discriminability measure based on an adaptive partitioning of the feature space according to the available class samples. It is intended to be used as a criterion in a classifier-independent feature selection procedure. The partitioning is performed according to a binary splitting rule and appropriate stopping criteria. Results from several tests withc Gaussian and non-GAussian, multidimensional and multicalss computer-generated samples, were very similar to those obtained using a Bayes error criterion function, i.e. the optimal feature subsets selected by both criterion functions were the same. The main advantage of the new measure is that it is computationally efficient.