ENFOQUE ASOCIATIVO PARA LA SELECCIÓN DE RASGOS

Abstract

In this thesis an Associative Memory Approach for Feature Selection is introduced. The pro- posed model constitutes a new model for data dimensionality reduction on pattern recogni- tion tasks. This model has its foundations in two major areas of pattern recognition theory, namely: Associative Memory and Bayesian Theory. On one side, the proposed model is based on the Hybrid Classi¯cation and Masking algorithm HCM and on the other it is based on likelihood ratio, taken from Bayesian Theory. This model shows competitive performance when compared against other important classi¯cation methods of the current literature.