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

BGonline.org Forums

XG training and NN question

Posted By: Timothy Chow
Date: Sunday, 1 August 2010, at 3:31 a.m.

In Response To: XG training and NN question (eXtreme Gammon)

I am curious as to whether you have any ideas as to how to address the Achilles heel of the current bots, namely the kinds of positions that arise from deep backgames. The bots misplay these positions so badly that it makes me wonder how much it affects their overall strength. I know you've said before that trying to get the NN to play these positions better seems to have the effect of making it play "normal" positions worse. However, perhaps there's some way of training two separate neural nets, the "normal" NN and the "deep backgame" NN, and then have the bot choose which NN to use (perhaps based on a third NN?).

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.