This is our first attempt to construct research (RL) agents. These are (hierarchically) driven by simple (hierarchical) entropy goals that allow the agent to be each time more informed about relations between concepts of some information input. Intially such an input is a collections of text documents.
iarroyof / herrl Goto Github PK
View Code? Open in Web Editor NEWHierarchical Entropy Relational Reinforcement Learning