LS-DRAUGHTS um sistema de aprendizagem de jogos de Damas baseado em algoritmos genéticos, redes neurais e diferenças temporais

• Ano de defesa: 2007
• Autor(a) principal: Castro Neto, Henrique de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; BR; Programa de Pós-graduação em Ciência da Computação; Ciências Exatas e da Terra; UFU
• Palavras-chave em Português: Aprendizagem automática; Aprendizagem de máquina; Aprendizagem incremental; Aprendizagem por reforço; Computação evolutiva; Algoritmo genético; Redes neurais; Busca minimax; Aprendizagem por diferenças temporais; Teoria dos jogos; Damas; Aprendizado do computador; Algoritmos genéticos; Inteligência artificial; Automatic learning; Machine learning; Incremental learning; Reinforcement learning; Evolutionary computation; Genetic algorithm; Neural network; Minimax search; Temporal diference learning; Game theory; Draughts; Checkers; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufu.br/handle/123456789/12575

This objective of this work is to propose a Learning System Draughts, LS-Draughts, that aims, through the Genetic Algorithm (GA), to generate automatically, a set of minimal features which are necessary and essential to a game of Draughts in order to optimizes the training of a player agent that learns to play Draughts. The learning method of this agent consists in approaching a Neural Network MLP through Reinforcement Learning TD(λ) along with the minimax search, with the mapping of the gam board NET-FEATUREMAP (done according to the features generated by the GA) and with the training process of self-play with cloning. The goal of the agent s self-learning, without any expert game analysis, is to allow an intelligent system to learn to play Draughts through its own experience. Such learning process is analogous to the process used by the NeuroDraughts system proposed by Mark Lynch. However, the LS-Draughts expands to NeuroDraughts as it generates automatically an effective and concise set of features of the NET-FEATUREMAP mapping, so that, the last one, uses a fixed and manually defined se of features. A tournament was promoted between the best player obtained by the LS-Draughts and the best available player of the LS-Draughts. The tournament s results, won by the player of the LS-Draughts, shows the fact that the GA represents and important improvement tool in these automatic players general performance.