Timezone: »

 
Poster
Agent57: Outperforming the Atari Human Benchmark
Adrià Puigdomenech Badia · Bilal Piot · Steven Kapturowski · Pablo Sprechmann · Oleksandr Vitvitskyi · Zhaohan Guo · Charles Blundell

Wed Jul 15 03:00 PM -- 03:45 PM & Thu Jul 16 02:00 AM -- 02:45 AM (PDT) @
in Poster Session 29 »

Atari games have been a long-standing benchmark in the reinforcement learning (RL) community for the past decade. This benchmark was proposed to test general competency of RL algorithms. Previous work has achieved good average performance by doing outstandingly well on many games of the set, but very poorly in several of the most challenging games. We propose Agent57, the first deep RL agent that outperforms the standard human benchmark on all 57 Atari games. To achieve this result, we train a neural network which parameterizes a family of policies ranging from very exploratory to purely exploitative. We propose an adaptive mechanism to choose which policy to prioritize throughout the training process. Additionally, we utilize a novel parameterization of the architecture that allows for more consistent and stable learning.

Keywords: Deep Reinforcement Learning · Reinforcement Learning · Reinforcement Learning - Deep RL

Author Information

Adrià Puigdomenech Badia (Deepmind)
Bilal Piot (DeepMind)
Steven Kapturowski (Deepmind)
Pablo Sprechmann (DeepMind)
Oleksandr Vitvitskyi (DeepMind)
Zhaohan Guo (DeepMind)
Charles Blundell (DeepMind)

More from the Same Authors