Q-Learning with Double Progressive Widening : Application to Robotics

Nataliya Sokolovska (1, 2) , Olivier Teytaud (1, 2) , Mario Milone (1)

1 LRI - Laboratoire de Recherche en Informatique
2 TAO - Machine Learning and Optimisation

Discretization of state and action spaces is a critical issue in $Q$-Learning. In our contribution, we propose a real-time adaptation of the discretization by the progressive widening technique which has been already used in bandit-based methods. Results are consistently converging to the optimum of the problem, without changing the parametrization for each new problem.

Domaines: Apprentissage [cs.LG]

Nataliya Sokolovska, Olivier Teytaud, Mario Milone. Q-Learning with Double Progressive Widening : Application to Robotics. ICONIP 2011, Nov 2011, China. pp.103-112. ⟨hal-00624832⟩