Título: Treinamento Condicionado Através de Esquemas de Transição
Autores: Uliana, Policarpo B.; Seara, Rui; Bermudez, José Carlos M.
Resumo: This paper proposes a new architecture for reinforcement learning of agents that are controlled via reward or punish stimuli. This proposal is based on transition schemes, which map the agent environment for controlling it. Such mapping uses cause and effect basic relations, which can be used to build complex control structures. This approach is an interesting alternative, leading to a new program strategy, which requires little information, as compared to conventional programming techniques.
Artigo em pdf: 4cbrn_087.pdf
Arquivo BibTex: 4cbrn_087.bib