Partially Observable Reinforcement Learning with Memory Traces

Onno Eberhard, Michael Muehlebach, Claire Vernade

Partially observable environments present a considerable computational challenge in reinforcement learning due to the need to consider long histories. Learning with a finite window…