"Near-Optimal Reinforcement Learning in Polynominal Time."

Michael J. Kearns, Satinder P. Singh (1998)
a service of Schloss Dagstuhl - Leibniz Center for Informatics