Rainbow: Combining Improvements in Deep Reinforcement Learning
2018; Association for the Advancement of Artificial Intelligence; Volume: 32; Issue: 1 Linguagem: Inglês
10.1609/aaai.v32i1.11796
ISSN2374-3468
AutoresMatteo Hessel, Joseph Modayil, Hado van Hasselt, Tom Schaul, Georg Ostrovski, Will Dabney, Dan Horgan, Bilal Piot, Mohammad Gheshlaghi Azar, David Silver,
Tópico(s)Modular Robots and Swarm Intelligence
ResumoThe deep reinforcement learning community has made several independent improvements to the DQN algorithm. However, it is unclear which of these extensions are complementary and can be fruitfully combined. This paper examines six extensions to the DQN algorithm and empirically studies their combination. Our experiments show that the combination provides state-of-the-art performance on the Atari 2600 benchmark, both in terms of data efficiency and final performance. We also provide results from a detailed ablation study that shows the contribution of each component to overall performance.
Referência(s)