We do train on 13x13 and extend to 19x19, but this was done in the game of Hex, because Hex was still too hard for computers -- whereas the excellent work by Deepmind has shown previously that humans can be beaten at Go. (https://www.facebook.com/notes/olivier-teytaud/optimization-machine-learning-artificial-intelligence-games-electricity/10162959845390472/ …)
PolyGame: An open-source framework for training AI players through self-play. Deals with many games, board size variation, partial observability... Interesting generalization tidbit: It plays Go on 19x19 at very good level after training only on 13x13.https://ai.facebook.com/blog/open-sourcing-polygames-a-new-framework-for-training-ai-bots-through-self-play/ …
-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Maybe my next Reinforcement Learning video should be about this.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Are you saying that the people claiming that DL was obviously limited as it was incapable of adapting to different Go board sizes were .... they were wrong ?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Wouls be interesting to see a simulated game of Go between
@DeepMind 's AlphaZero and@facebookai 's PolyGame?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.