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

BGonline.org Forums

XG Update (Crossposted from BG News FB group)

Posted By: Frank Berger
Date: Sunday, 16 July 2023, at 11:12 a.m.

In Response To: XG Update (Crossposted from BG News FB group) (Bob Koca)

I think that could be done, just a trivial case:

- I post process the NN output if a checker is off gammon prob is zero (there might be some 0,000002 or so) or when the move doesn't make any difference I add a tiny amount to reward taking a checker off, so avoiding "inhuman" plays. This are trivial and harmless cases.

- I believe it is quite easy to have a rule based approach to things like the Kauder paradox or this position here (not for backgames). But that has 2 IMHO large drawbacks: You don't have probabilities anymore (or at most a best guess) and at the borders the evaluation function might have quirks (IIRC Berliner wrote about the border issues)

- I still believe that self learning is able to solve Kauder and deep backgames (when it works in Go it should work here too). I think the right tweaking wasn't found yet (and how many persons working on it, I guess that is the main culprit). When the weather get's cooler in autumn I make the next attempt.

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.