It's not available online, but I will send it if someone ask me to.
(in private e-mail)
Lukasz
2009/10/13 Petr Baudis :
> On Fri, Sep 25, 2009 at 11:51:21AM +0200, Łukasz Lew wrote:
>> I tried CRAVE in my master thesis 4 years ago. The context was a
>> growing decision tree.
>> It didn't work as w
On Fri, Sep 25, 2009 at 11:51:21AM +0200, Łukasz Lew wrote:
> I tried CRAVE in my master thesis 4 years ago. The context was a
> growing decision tree.
> It didn't work as well.
This sounds similar to one idea of mine; is your thesis available
anywhere? (Or any other material published on CRAVE.)
http://users.soe.ucsc.edu/~dph/mypubs/AMAFpaperWithRef.pdf
Cheers,
David
On 25, Sep 2009, at 12:34 PM, Peter Drake wrote:
Yes. I believe Fuego does this. See also Helmbold and Parker-Wood,
"All-Moves-As-First Heuristics in Monte-Carlo Go":
(Does anyone have a URL for this one? I can't see
Here's a suggestion to extend RAVE to better handle it:
There are 20 points within keima distance of any point not close to the
edge.(5*5 without the corners)
When RAVE values are backed up, they are put into the category defined by
the previous opponents move.
(21 categories, 20 + other. All adde
> Stefan Kaitschick wrote:
> >Here's a suggestion to extend RAVE to better handle it:
> >There are 20 points within keima distance of any point not close to the
> >edge.(5*5 without the corners)
> >When RAVE values are backed up, they are put into the category defined by
> >the previous opponents m
Stefan Kaitschick wrote:
>Here's a suggestion to extend RAVE to better handle it:
>There are 20 points within keima distance of any point not close to the
>edge.(5*5 without the corners)
>When RAVE values are backed up, they are put into the category defined by
>the previous opponents move.
>(21
MCTS, even though it walks to the end of the earth, has it's own horizon
effect.
The name is more fitting for depth-limited alpha-beta search ofcourse.
It's a kind of procrastination. Finding a lot of useless things to do before
admitting
an undesirable, but unavoidable consequence.
Even if a f
Stefan Kaitschick wrote:
About the move order problem: I'm skeptical about finding an efficient
full
board algorithm that detects the consequences of move order.
In this context, "the order of moves" does not mean a long sequence.
It means something like "If I already played A, B is preferab
Brian Sheppard wrote:
Fuego uses a lower weight for distant moves than for nearby moves.
I suspect that isn't much better than using uniform weight. I am
hope that Martin or Markus will comment.
I measured a winning rate of 55.1(+-0.8)% of Fuego with weighted RAVE
updates vs. the version w
On Sat, Sep 26, 2009 at 6:11 AM, Brian Sheppard wrote:
> In my dream, I went on to overcome these difficulties and solve the problem.
> But when I woke up I couldn't remember how I did it...
That's too bad. I get almost all my good ideas in my sleep. I think
because the brain is more relaxed and
Stefan Kaitschick wrote:
>About the move order problem: I'm skeptical about finding an efficient full
>board algorithm that detects the consequences of move order.
In this context, "the order of moves" does not mean a long sequence.
It means something like "If I already played A, B is preferable
I toyed with that a while back and the results were disappointing.
Every move's best reply was to tenuki to at a big point. You may need
to subtract out the normal rave values to find which replies become
more important than normal, even if they aren't the biggest moves
overall. Or maybe on
ad end. (I am fairly disenchanted with using patterns as context, though.)
-?Dave Hillis
-Original Message-
From: Peter Drake
To: computer-go
Sent: Fri, Sep 25, 2009 3:41 pm
Subject: Re: [computer-go] Generalizing RAVE
On Sep 24, 2009, at 8:45 PM, terry mcintyre wrote:
Indeed it i
On Sep 24, 2009, at 8:45 PM, terry mcintyre wrote:
Indeed it is. How may a program reason about the order of moves? At
higher levels of play, the order of moves is often crucial.
I plan to try the following:
Store win and run counts for each move in the context of the two
previous moves.
Yes. I believe Fuego does this. See also Helmbold and Parker-Wood,
"All-Moves-As-First Heuristics in Monte-Carlo Go":
(Does anyone have a URL for this one? I can't seem to find it online,
but I have a paper copy in front of me.)
Peter Drake
http://www.lclark.edu/~drake/
On Sep 25, 2009,
Yup, I tried something like that, too, with a similar lack of luck.
Peter Drake
http://www.lclark.edu/~drake/
On Sep 25, 2009, at 6:39 AM, Brian Sheppard wrote:
I have another way to fail to improve on RAVE. :-)
I tested a method that gives higher weight to recent RAVE data. The
method
us
It is exactly the same as my thought. I also have tried CRAVE, but the
results were worse than normal RAVE.
While RAVE is a very efficient algorithm, it strongly limits scalability
of the program. It typically makes a fatal mistake in the position that
the order of moves are important. We definit
From: Brian Sheppard
> I have another way to fail to improve on RAVE. :-)
Well, that's great news.
Thomas Edison was once asked if he felt discouraged by 10 thousand failed
experiments, and he said "Not at all; I now know ten thousand ways to not build
an
I tried CRAVE in my master thesis 4 years ago. The context was a
growing decision tree.
It didn't work as well.
Lukasz
On Fri, Sep 25, 2009 at 06:19, David Fotland wrote:
> Tried CRAVE also, using 3x3 patterns as the context. It didn't work.
>
> David
>
>> -Original Message-
>> From: c
Tried CRAVE also, using 3x3 patterns as the context. It didn't work.
David
> -Original Message-
> From: computer-go-boun...@computer-go.org [mailto:computer-go-
> boun...@computer-go.org] On Behalf Of Peter Drake
> Sent: Thursday, September 24, 2009 12:00 PM
> To: Computer Go
> Subject:
Peter Drake wrote:
>The more I study this and try different variants, the more impressed I
>am by RAVE. "Boards after the current board" is a very clever way of
>defining similarity. Also, recorded RAVE playouts, being stored in
>each node, expire in an ele
Peter Drake wrote:
>The more I study this and try different variants, the more impressed I
>am by RAVE. "Boards after the current board" is a very clever way of
>defining similarity. Also, recorded RAVE playouts, being stored in
>each node, expire in an elegant way. It still seems that RAVE f
22 matches
Mail list logo
