| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.39 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
De entre todos os paradigmas de aprendizagem actualmente identificados, a Aprendizagem por Reforço revela-se de especial interesse e aplicabilidade nos inúmeros processos que nos rodeiam: desde a solitária sonda que explora o planeta mais remoto, passando pelo programa especialista que aprende a apoiar a decisão médica pela experiencia adquirida, até ao cão de brincar que faz as delícias da criança interagindo com ela e adaptando-se aos seus gostos, e todo um novo mundo que nos rodeia e apela crescentemente a que façamos mais e melhor nesta área.
Desde o aparecimento do conceito de aprendizagem por reforço, diferentes métodos tem sido propostos para a sua concretização, cada um deles abordando aspectos específicos. Duas vertentes distintas, mas complementares entre si, apresentam-se como características chave do processo de aprendizagem por reforço: a obtenção de experiência através da exploração do espaço de estados e o aproveitamento do conhecimento obtido através dessa mesma experiência.
Esta dissertação propõe-se seleccionar alguns dos métodos propostos mais promissores de ambas as vertentes de exploração e aproveitamento, efectuar uma implementação de cada um destes sobre uma plataforma modular que permita a simulação do uso de agentes inteligentes e, através da sua aplicação na resolução de diferentes configurações de ambientes padrão, gerar estatísticas funcionais que permitam inferir conclusões que retractem entre outros aspectos a sua eficiência e eficácia comparativas em condições específicas.
Of all the currently identified learningparadigms, Reinforcement Learningappears to be of special interest and applicability in the numerous processes that surround us: from the lonely probeexploring the most remote planet, through the specialist programthat learns to support medical decision-making through gained experience, to the toy dogthat delights the child interacting with it and adapting to its liking, it's a whole new world surrounding us beckoning us to do more and better in this area. Since the reinforcement learning concept appeared, different methods have been proposed for its achievement, each of them approaching specific aspects. Two distinct, but complementary, strands appear as key features in the reinforcement learning process: getting experience through the explorationof state space and the exploitationof the knowledge gained through this same experience. This thesis proposes to select some of the current most representative methods, in both explorationand exploitationstrands, implement each of these on a modular platform that allows the simulation of intelligent agents and, through its application in solving different configurations of standard environments, generate functional statistics that allow conclusions to be inferred that reflect, amongst others, their comparative efficiencyand effectivenessin specific conditions.
Of all the currently identified learningparadigms, Reinforcement Learningappears to be of special interest and applicability in the numerous processes that surround us: from the lonely probeexploring the most remote planet, through the specialist programthat learns to support medical decision-making through gained experience, to the toy dogthat delights the child interacting with it and adapting to its liking, it's a whole new world surrounding us beckoning us to do more and better in this area. Since the reinforcement learning concept appeared, different methods have been proposed for its achievement, each of them approaching specific aspects. Two distinct, but complementary, strands appear as key features in the reinforcement learning process: getting experience through the explorationof state space and the exploitationof the knowledge gained through this same experience. This thesis proposes to select some of the current most representative methods, in both explorationand exploitationstrands, implement each of these on a modular platform that allows the simulation of intelligent agents and, through its application in solving different configurations of standard environments, generate functional statistics that allow conclusions to be inferred that reflect, amongst others, their comparative efficiencyand effectivenessin specific conditions.
Description
Keywords
Agentes inteligentes Processos de decisão de Markov Aprendizagem por reforço Política avaliativa Política comportamental Intelligent agentes Markov decision processes Reinforcement learning Evaluation policy Behaviour policy
Pedagogical Context
Citation
Pessoa, João Manuel Dionísio - Análise funcional comparativa de algoritmos de aprendizagem por reforço. Lisboa: Instituto Superior de Engenharia de Lisboa, 2011. Dissertação de mestrado.