deep reinforcement learning (DRL) has achieved remarkable success in
sequential decision-making problems. However, existing DRL agents make
decisions in an opaque fashion, hindering the user from establishing trust and
scrutinizing weaknesses of the agents. While recent research has de