On Apr 21, 2009, at 11:57 AM, [email protected] wrote:
>Not using a new login after significant changes leads to some issue
>with both rating schrmes. It helps you and everyone else.
Pebbles learns from every game it plays. So I can't agree; drift is
inherent.
Interesting. Is this an opening book or something else such as
successful patterns? Do you take the opponent's rank into account when
learning?
>Do you mind sharing what pebbles does? UCT+RAVE? Any other
enhancements?
It is UCT+RAVE, but I am not sure that what I have implemented is
the same as what others have. For example, I have never been
clear on how RAVE differs from AMAF. I have done a lot of reading,
but there are a lot of overlapping ideas and variations and
alternatives.
Not to mention tweaks and tunings.
AMAF and RAVE are the same thing. The MoGo team pioneered use of AMAF
but called it RAVE because of their paper's target audience.
Aside from the learning mentioned above, the engine uses only
methods that have been published in this mailing list or in Mogo
papers.
And then only a small subset of those, and I am not sure that I have
always carried out the intent of the inventors.
I'm sure we all feel that way :)
Brian
The Average US Credit Score is 692. See Yours in Just 2 Easy Steps!
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
