Accelerating Self-Play Learning in Go
Rémi Coulom <[email protected]> Sun, 24 Mar 2019 15:50:18 +0100
| Newsgroups | gmane.games.devel.go |
|---|---|
| Message-ID | <CANs0NNVDU_oY-wOL=qdG6m=jr5DuAzNxg7y4UnTkCpePWUiWkg@mail.gmail.com> |
Hi, I have just found out that the list is not sending emails to my free.fr email address any more. So I subscribed with my gmail address, which I hope should work better. I had missed that very interesting message by David Wu ( http://computer-go.org/pipermail/computer-go/2019-March/010991.html). I simply wish to share my confirmation that maximizing territory works well. I have been training Crazy Stone to maximize territory since the beginning. That's how it reached the top of CGOS after a couple months of training with only 2-3 (Volta) GPUs in February, 2018. It took Leela several more months to reach a similar strength, with considerably more computing power. Rémi _______________________________________________ Computer-go mailing list [email protected] http://computer-go.org/mailman/listinfo/computer-go