Sample-Efficient Deep Reinforcement Learning via Episodic Backward Update
This addresses the problem of high sample requirements for reinforcement learning practitioners, offering a significant reduction in data needs, though it is an incremental improvement over existing methods.
The paper tackles sample inefficiency in deep reinforcement learning by introducing Episodic Backward Update (EBU), which samples whole episodes for direct value propagation, achieving the same performance as DQN with only 5-10% of samples in Atari 2600 games.
We propose Episodic Backward Update (EBU) - a novel deep reinforcement learning algorithm with a direct value propagation. In contrast to the conventional use of the experience replay with uniform random sampling, our agent samples a whole episode and successively propagates the value of a state to its previous states. Our computationally efficient recursive algorithm allows sparse and delayed rewards to propagate directly through all transitions of the sampled episode. We theoretically prove the convergence of the EBU method and experimentally demonstrate its performance in both deterministic and stochastic environments. Especially in 49 games of Atari 2600 domain, EBU achieves the same mean and median human normalized performance of DQN by using only 5% and 10% of samples, respectively.