IRL baselines for gfootball?

15 views

Skip to first unread message

Stephen Casper

unread,

Dec 24, 2020, 8:52:43 PM12/24/20

to Google Research Football

Hi all, I am working on a project that involves inverse reinforcement learning for GRF. I'm wondering if anyone has had more success than I have with it. Right now, I am able to train imitators to agree with demonstrators around 27% of the time using an implementation of GAIL. I'm using extracted representations and action set v1.

Thanks in advance.

Cheers,

Reply all

Reply to author

Forward

0 new messages