First-Order Tree-Type Dependence between Variables and Classification Performance

Authors:
Sarunas Raudys;Ausra Saudargiene
Affiliations:
Institute of Mathematics and Informatics, Akademijos, Lithuania;Institute of Mathematics and Informatics, Akademijos, Lithuania
Venue:
IEEE Transactions on Pattern Analysis and Machine Intelligence
Year:
2001

Citing 4
Cited 3

Introduction to statistical pattern recognition (2nd ed.)

Introduction to statistical pattern recognition (2nd ed.)
Small Sample Size Effects in Statistical Pattern Recognition: Recommendations for Practitioners

IEEE Transactions on Pattern Analysis and Machine Intelligence
Evolution and generalization of a single neurone: I. single-layer perceptron as seven statistical classifiers

Neural Networks
Structures of the Covariance Matrices in the Classifier Design

SSPR '98/SPR '98 Proceedings of the Joint IAPR International Workshops on Advances in Pattern Recognition

An Algebraic Formalism for Computing the Moments of Distributions of Quadratic Forms

Automation and Remote Control
Evolution of Multi-class Single Layer Perceptron

ICANNGA '07 Proceedings of the 8th international conference on Adaptive and Natural Computing Algorithms, Part II
A pool of classifiers by SLP: a multi-class case

ICIAR'06 Proceedings of the Third international conference on Image Analysis and Recognition - Volume Part II

Quantified Score

Hi-index	0.14

Visualization

Abstract

Structuralization of the covariance matrix reduces the number of parameters to be estimated from the training data and does not affect an increase in the generalization error asymptotically as both the number of dimensions and training sample size grow. A method to benefit from approximately correct assumptions about the first order tree dependence between components of the feature vector is proposed. We use a structured estimate of the covariance matrix to decorrelate and scale the data and to train a single-layer perceptron in the transformed feature space. We show that training the perceptron can reduce negative effects of inexact a priori information. Experiments performed with 13 artificial and 10 real world data sets show that the first-order tree-type dependence model is the most preferable one out of two dozen of the covariance matrix structures investigated.