The project is organized along three main research work packages.
Work package 1: Transfer of exploration-exploitation strategies
Task T1.1: Knowledge transfer in multi-armed bandit.
Task T1.2: Towards exploration-exploitation transfer in RL.
Work package 2: Transfer solutions for approximated reinforcement learning
Task T2.1: Sample complexity reduction
Task T2.2: Automatic feature generation
Work package 3: Hierarchical transfer reinforcement learning
Task T3.1: Automatic construction of skills
Task T3.2: Hierarchical approximation schemes
Task T3.3: Integration and testing in complex environments
