On the stability of reinforcement learning under partial observability and generalizing representations (2010)
[BibTex]