Toward a data efficient neural actor-critic

Matthieu Zimmer (1, 2) , Yann Boniface (2) , Alain Dutech (1)

1 MAIA - Autonomous intelligent machine
2 UL - Université de Lorraine

A new off-policy, offline, model-free, actor-critic reinforcement learning algorithm dealing with continuous environments in both states and actions is presented. It addresses discrete time problems where the goal is to maximize the discounted sum of rewards using stationary policies. Our algorithm allows to trade-off between data-efficiency and scalability. The amount of a priori knowledge is kept low by: (1) using neural networks to learn both the critic and the actor, (2) not relying on initial trajectories provided by an expert, and (3) not depending on known goal states. Experimental results show better data-efficiency than 4 state-of-the-art algorithms on two benchmark environments.

Mots clés: Continuous Spaces; Actor-Critic; Neural Networks

Domaines: Intelligence artificielle [cs.AI]

Matthieu Zimmer, Yann Boniface, Alain Dutech. Toward a data efficient neural actor-critic. EWRL 2016 - The 13th European Workshop on Reinforcement Learning, Dec 2016, Barcelona, Spain. ⟨hal-01413885⟩