R.Schoknecht. A necessary condition of convergence for reinforcement learning with function approximation (2002)

by A Merke
Venue:Proceedings of ICML