Mark B. Ring

Better Generalization with Forecasts

Q-Error as a Selection Mechanism in Modular Reinforcement-Learning Systems

Using Predictive Representations to Improve Generalization in Reinforcement Learning