Sketch Decompositions for Classical Planning via Deep Reinforcement Learning

Michael Aichmüller, Hector Geffner

In planning and reinforcement learning, the identification of common subgoal structures across problems is important when goals are to be achieved over long horizons. Recently, it…