"Proto-value functions: developmental reinforcement learning."

Sridhar Mahadevan (2005)
a service of  Schloss Dagstuhl - Leibniz Center for Informatics