AlphaGo Zero is great, but hold on: self-play is one of the oldest ideas in ML, and humans take far less than 5 million games to master Go.
-
Pedro, how generalizable is this result? How many (real-world) problems/ situations are amenable to self play?
Not many, a priori. But if you can frame them as turn-taking games with many turns and a large space of moves, perhaps.
-
-
Many people are training robot controllers using simulations. But the added challenge is to transfer from sim to real world.
-
Some groups are starting to demonstrate success with this. e.g.https://arxiv.org/abs/1707.07907
- Näytä vastaukset
Uusi keskustelu -
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.