A parsimonious generation of combinatorial neural model

This paper presents a new approach to reduce the space problem due to combinatorial explosion of CNM (Combinatorial Neural Model) method. First we show a description of CNM, proposed by Machado and Rocha [MAC 91], [MAC 92], [MAC 92a], [MAC 97], as a variation of fuzzy neural network introduced as an alternative to meet many requirements, such as expressiveness, inteligibility, plasticity and flexibility. Our approach represents an alternative to generate the CNM network with certainty factors for each hypothesis. We demonstrate by means of a simple practical example that the number of combinations can be really reduced. Sistemas Inteligentes Red de Universidades con Carreras en Informática (RedUNCI)