Discovering symbolic policies with deep reinforcement learning

    Mikel Landajuela * 1 Brenden K. Petersen * 1 Sookyung Kim * 1 Claudio P. Santiago 1 Ruben Glatt 1
              T. Nathan Mundhenk 1 Jacob F. Pettit 1 Daniel M. Faissol 1

             Abstract
   Deep reinforcement learning (DRL) has proven
   successful for many difficult control problems by
   learning policies represented by neural networks.
   However, the complexity of neural network-based
   policies—involving thousands o ...