Algoritmos Básicos de la Computación Cognitiva para la Programación de Cerebros Artificiales

Abstract

Scientific reductionism finds solutions that are almost perfect, inflexible, and specific for a big variety of problems. In general, these solutions are models of causal deduction that go from causes to effects. This thesis presents 3 intelligent algorithms of causal induction that go from effects to causes: The mental image reconstructor, the evolutionary characterizer, and the hierarchical causal router. These algorithms use networks of interrelated nodes that exploit the pattern space to find the best causes or solutions that explain the effects or problems presented as evidence. For each algorithm, an application is presented. But these algorithms are very general and applicable to many problems. The mental image reconstructor solves the minesweeper videogame in few seconds. The evolutionary characterizer infers the tridimensionality of bidimensional photos through a basic system of artificial vision. And the hierarchical causal router is presented as a theoretical solution for learning and self-organizing in real time the causal geometries of a robotic arm and of every type of robot. But this problem is a frontier of science. It is not solved here but some strategies to solve it are suggested.