[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Make your guesses - Answers

Posted By: Frank Berger
Date: Wednesday, 13 August 2025, at 8:21 p.m.

In Response To: Make your guesses - Answers (Bob Koca)

The problem is that I only know very little about XGs architecture. Once Xavier said here that XG has "a lot of" nets (whatever a lot is. 5? 7? 9? ...) and I know nothing about the training method. I assume that supervized training is used, maybe in combination with reinforment learning but that is just guessing.

I use only reinforcement learning (i.e. learning by self play) which seems to be more robust evaluating exotic positions like those extreme containment games. In the BG community it was said that RL leads only to intermediate level of play, but what Deepmind did (e.g. AlphaGo, AlphaZero, MuZero) indicates that this is wrong.

What astonished me, is that I thought only snakes would be an issue for XG, but this position shows this is not true (in the few manuals rollouts I did there was no snake at all).

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.