Rohin Shah
TalkRL: The Reinforcement Learning Podcast
00:00
How to Get Robustness to Environments in Collaborative Games
The way you do it is you just say, all right, we're going to make the environment a our competitor. The environment's going to like, try and like make itself super complicated. And so then this makes sure that you have to be robust to whichever environment you find yourself. This doesn't work great if you like literally make it adversary, because sometimes it many like interesting collaborative games. If your partner is an adversary, they can just guarantee that you get minimum reward. It's often not difficult. But there are papers that try to do this sort of thing. I think many of them did end up following a both the self play work and this paper of ours.
Transcript
Play full episode
