Exploration in Approximate Hyper-State Space
                 for Meta Reinforcement Learning

          Luisa Zintgraf 1 Leo Feng 2 Cong Lu 1 Maximilian Igl 1 Kristian Hartikainen 1
                    Katja Hofmann 3 Shimon Whiteson 1

             Abstract
   To rapidly learn a new task, it is often essential
   for agents to explore efficiently – especially when
   performance matters from the first timestep. One
   way to learn such behaviour is via meta-learning.
   Many existing methods h ...