INTELIGÊNCIA COMPUTACIONAL APLICADA EM MACHINE LEARNING / COMPUTATIONAL INTELLIGENCE APLPLIED IN THE FIELD OF MACHINE LEARNING

AUTOR(ES)
DATA DE PUBLICAÇÃO

1999

RESUMO

This work investigates the application and performance of the Computational intelligence technics in the field of Machine Learning. In particular, was investigated the application of intelligent systems (Neural Networks, genetic Algorithms and fuzzy Logic) in the development of algorithms that codify inductive mechanisms in Machine Learning. This work was developed in two main steps: a research of Machine Learning bibliography, and the development of three intelligent systems: neural Networks applied to the game of backgammon, Genetic Algorithms in the evolution of an autonomous robot control system, and fuzzy Logic applied to robot control. The bibliography research involved looking for technical literature about Machine Learning and Computational Intelligence. Were used in this research books specialized in the area, and technical papers about the themes treated in this dissertation. The modeling of the backgammon learning algorithm, based on Neural Networks, was implemented using a reinforcement learning method known as TD(l), which operate by the principle of trial and error, giving a reward for actions that brings a good result. The game of backgammon was chosen because of its huge number of possible situations that can be faced during the game, due to estocastic factor (dice) attached to the game, making a traditional approach very difficult, if not even inefficient. The evolution of the autonomous robot control system using Genetic Algorithms is inspired in the evolution of a behavior pattern of behavior to deal with the faced at each moment, in order to achieve a goal in an optimal or sub-optimal way. The robot control system using Fuzzy Logic demonstrates the potential of this technic to control problems, codifying through fuzzy rules, which are similar to the human way to in its environment. The results presented demonstrate the potential of the Computational Intelligence technics, inspired in biology and nature, in the field of Machine Learning, showing through the examples implemented the knowledge acquisition capacity by experience, using inductive mechanisms instead of programming specific solutions for the problems presented

ASSUNTO(S)

fuzzy logic algoritmos geneticos neural networks logica fuzzy aprendizado de computador genetic algorithms redes neurais learning for computer

Documentos Relacionados