Hierarchical learning from human preferences and curiosity

Applied Intelligence - Recent success in scaling deep reinforcement algorithms (DRL) to complex problems has been driven by well-designed extrinsic rewards, which limits their applicability to many...