Characterizing the benefits of model-based vs . direct-control learning in exploration

@inproceedings{Schuurmans2007CharacterizingTB,
  title={Characterizing the benefits of model-based vs . direct-control learning in exploration},
  author={Dale Schuurmans},
  year={2007}
}
We consider the problem of learning a good control function f r m reinforcement. Although much studied, the issue of how (or even whether) to u se prior knowledge to scale reinforcement learning up to real-world problems remains a important open issue. We investigate the inherent data-complexity of reinforcem ent learning when prior knowledge is expressed… CONTINUE READING