Optimistic initialization and greediness lead to polynomial time learning in factored MDPs.

Istvan Szita, Andrs Lrincz

Browse the full ICML paper archive.