Neural learning combined with game-theoretic reasoning – check out our AAAI 2021 paper: “Learning game-theoretic models of multiagent trajectories using implicit layers”.