search

Kaixhin / Rainbow

Rainbow: Combining Improvements in Deep Reinforcement Learning
MIT License
1.59k stars 284 forks source link

--learn-start #67

Closed zyzhang1130 closed 4 years ago

zyzhang1130 commented 4 years ago

Hi, May i check with you why there are 20000 step before training by default? What is the use of it? Thank you.

Kaixhin commented 4 years ago

This allows a sufficient amount of data to be gathered before training, to prevent overfitting to a small dataset.

zyzhang1130 commented 4 years ago

noted with thanks