In this paper we explore how application-users can, in an interactive way, test properties about the state of GUI applications that can be classified as local state transition sys...
We present a new subgoal-based method for automatically creating useful skills in reinforcement learning. Our method identifies subgoals by partitioning local state transition gra...