Reinforcement Learning in POMDP's via Direct Gradient Ascent.

Jonathan Baxter, Peter L. Bartlett

Browse the full ICML paper archive.