Sobre las redes neuronales y la computación paralela

Abstract

Neurocybernetic, Systems Theory and Perception, and Artificial lntelligence Systems, interact in such a way that the methods and procedures of one of them provide clues in order to face and resolve problems in the others. This paper shows the interaction among the algebraic-analytical data fields theory, the complete descriptions that may be truncated for practical visual tasks and the structural-functional characteristics of neural nets. In this way, on researching the properties and requirements that complete descriptions had to verify, we may conclude that a general characteristic of transformations combining partitions and functionals, is the esentially parallel character of the computational structure that performs it. Then, the temporal factor, which had been forgotten in the former algebraic-analytical formulation has been introduced here, by considering layered and neural nets. This consideration allows's to clarify the reach and limitations of fast computational algorithms.