Análise e Implementação de Algoritmos para a Aprendizagem por Reforço

• Ano de defesa: 2014
• Autor(a) principal: Medeiros, Thiago Rodrigues
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraí­ba; BR; Informática; Programa de Pós Graduação em Informática; UFPB
• Palavras-chave em Português: Aprendizado de Máquina; Aprendizado por Reforço; Análise; Biblioteca; Machine Learning; Reinforcement Learning; Library; Statistical Analysis; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/tede/6119

The Reinforcement Learning is a subfield of machine learning and can be defined as a learning problem. An intelligent system that faces this problem, understands from rewards if the actions you are performing in the environment are good or bad. There are several methods and algorithms found in the literature to solve the problems of reinforcement learning. However, each of them have their advantages and disadvantages. From this, this paper presents a statistical analysis of some algorithms and a library of reinforcement learning, called AILibrary-RL. The AILibrary-RL is a library that has the objective to facilitate, organize and promote reusability of code, to implement systems that have this kind of problem. Before its development, a bibliographic survey of the main methods that solve this problem, aimed at statistical analysis of the data was performed in order to evaluate its advantages and disadvantages in different environments. This dissertation described the whole process of this work, since the survey bibliographic, analysis of the methods, mechanisms and library construction.