[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Bug-gnubg] Some questions & comment about contacting me
From: |
Thomas Hauk |
Subject: |
[Bug-gnubg] Some questions & comment about contacting me |
Date: |
Wed, 26 Nov 2003 00:00:21 -0700 (MST) |
Sorry Achim, I missed your question:
> Just one question: What is this thesis for?
mainly because I do not subscribe to the gnubg mailing list, just because
there's too much discussion about stuff that isn't code-related for me to
really care about. That doesn't mean I don't want to answer your questions
(although you did mention something contrary to that to my supervisors in
Graz!), just that you'll have to remember to CC: me at
<address@hidden> so I can respond. 8)
My thesis is investigating an improvement to expectimax search (for game
trees that include chance nodes, such as dice rolls, or dealing cards)
called *-minimax, developed by Bruce Ballard in 1983. The main focus was
on backgammon, since relatively little improvement has been done to
backgammon search, compared to evaluation functions (TD-trained NNs) for
example.
At this point we haven't seen a great deal of improvement in deep search
-- perhaps no improvement, really -- but I am continuing my work
nonetheless.
Have any of you done any benchmarking, playing gnubg versus itself, using
different ply settings? Say, gnubg at 0-ply versus gnubg at 1-ply or
2-ply? Have you seen statistically significant differences?
I am running DMP games right now, as my search algorithms do not yet
handle the cube. Any thoughts on that?
--
Some people, when confronted with a problem, think "I know, I'll use
regular expressions." Now they have two problems.
--Jamie Zawinski, in comp.lang.emacs
- [Bug-gnubg] Some questions & comment about contacting me,
Thomas Hauk <=