IDENTIFICATION OF RELEVANT SIGNAL FEATURES

Wojciech Cholewa

Abstract

The paper deals with design of diagnostic classifiers. The main goal is to illustrate an original way of identification of useful signal features on the basis of learning data prepared as a set of examples. One indicated some possibilities of application of a criterion based on the expectation that results of unsupervised clustering in a new limited space should be compatible with results of classification of learning data.

Keywords

diagnostic classifier, feature selection

Download

IMEKO-TC10-2005-01.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings