richardbaihe / paperreading

NLP papers
MIT License
2 stars 0 forks source link

ICLR 2021 | MASTERING ATARI WITH DISCRETE WORLD MODELS #62

Closed richardbaihe closed 3 years ago

richardbaihe commented 3 years ago

This paper investigate model-based method for Atari benchmarks. They show their method can outperform the model-free method significantly by learning a world model first and learning a behavior model later.

image image