Episode 22: Archit Sharma, Stanford, on unsupervised and autonomous reinforcement learning
Generally Intelligent
00:00
Optimize Infinite Horizon Discounting?
There's always not something that sort of intuitively brought me the wrong way out to like optimize infinite horizon discounting. The first one is really about maximizing the information in your MDP. And the second one is just about getting things done. So those in practice can end up being very, very different objectives. I mean, your other anywhere for relations seems somewhat better.
Transcript
Play full episode
