It is a great article overall. I would like it more if it mentions Mogo, at
least Follow from the opponent's previous move was actually Mogo's
invention in the famous UCT paper, not Fuego's, not to mention a lot of
Mogo's achievements on 9x9. But I really like the paragraph describing the
We had no control over the title page art. The IEEE art dept gave the following
explanation for their work:
The opening art for the story is conceptual and illustrative. It is not meant
to show a game in progress, but rather the idea of AI for Go. We took the game
elements (board, stones) and
Crazy Stone won the first game against Yoda. Now Zen is playing.
I will send updates on
http://computer-go-adventures.blogspot.ca
Official page:
http://entcog.c.ooco.jp/entcog/densei/eng/index.html
Martin
___
Computer-go mailing list
Gogui can be run with option -Xmx for large files, see:
http://gogui.sourceforge.net/doc/bugs.html#idp7970336
Martin
___
Computer-go mailing list
Computer-go@dvandva.org
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
Thank you Detlef for doing these tests!
I want to get more people interested into this scaling, therefore I did
also some scaling tests fuego against pachi :)
It is not as bad as oakfoam against pachi, but pachi scales a lot better
than fuego too. (attached file) To avoid additional
Are my fuego parameters ok?
opponent_settings2='uct_param_player ignore_clock 1\nuct_param_player
max_games '+str(playouts)+'\nuct_param_player resign_min_games 5000
\nuct_param_search number_threads 8\nuct_max_memory 80
\nuct_param_player reuse_subtree 1'
Hi Detlef,
looks OK
Bent four is hard for me. Is this right?
Without tree search, right playout result is maybe
If many ko threats are left, and white to move.
W lib is 4, W is 0% live.
W lib is 5, W is 66% live. white has to play J8.
W lib = 6, W is 66% live.
If there is no ko threat, and white to
Thank you Don for all your contributions to computer Go!
All the best for Komodo, and for your health.
Martin
P.S: Everyone, you can see Don's Komodo chess program in action in the ongoing
TCEC tournament.
http://tcec.chessdom.com/live.php
It would be cool to have such an event for Go
The tournament will be this weekend. 13×13, Chinese rules, komi 7½.
See
http://www.gokgs.com/tournInfo.jsp?id=799
I am sending this on behalf of Nick who is away from his computer at the
moment. But please still register with him - he will be back before the
tournament starts.
Martin
I want to support Japanese rules in Fuego for UEC cup (and for human opponents
who prefer these rules). I looked at test cases and previous discussions on the
list. My summary so far is on
http://sourceforge.net/apps/trac/fuego/wiki/JapaneseRules
I did not find any discussions younger than
I wrote a technical report mainly on this issue a few years ago. It has many
pictures that explain what can go wrong - at least with Fuego, but similar
problems seem to happen with other programs too.
https://www.cs.ualberta.ca/research/theses-publications/technical-reports/2010/TR10-08
The 29 stone game wasn't really a demonstration game.
My memory is that I had just finished igowin 9x9, and Martin played through
all its levels up to professional, without losing a game. I was surprised,
and while we were talking the subject of a 29 stone game on 19x19 came up.
I
I think that Martin - when he plays - will be beaten by a top bot within
the next two years, even at slow time controls.
Hi Ingo,
is this an official challenge, or an off-hand comment?
Martin___
Computer-go mailing list
Does anyone here know which rules will be used for the Olympiad? Especially
regarding:
- komi
- time limits
- remote play
thanks
Martin
___
Computer-go mailing list
Computer-go@dvandva.org
The main limitation of Fuego, and I suspect all other MCTS programs, is in the
playouts, not in the search. I wrote up my impressions and some analysis after
the matches vs strong humans in Barcelona last year.
* Do you do this for 19x19 only, or also for 9x9 and 13x13?
* If so, do you have other ranges (instead of 0.98 - 1.00 ...)
for the smaller board sizes?
* Did you observe, if these modifications in evaluation interfer with
dynamic/linear komi in some way?
Same for all board sizes. And
Is there a good reason, why one should not take into account height
(number of stones) of the win?
It feels like loosing information to me?!
In Fuego, we give a small bonus for the size of the win/loss. So a win is worth
between 0.98 and 1.00 depending on the size of the win, and a loss is
5x1000 games, no opening book, Fuego svn version 1467.
Number of simulations varied from 3K (3000) per move to 300K.
I was interested in whether the recent fix to allow Fuego to play for draws
with integer komi worked, and whether the number of draws increases with longer
searches.
Here are the
Thank you Magnus for the test case. I am not sure what is the cause but I hope
we can fix it.
Martin
Hi, I just observed a surprising loss on CGOS with Fuego vs Amigo. The
game is clearly won for Fuego, but in the end just before the last
dead group is captured Amigo captures a
Fuego 1.0 is available for download at
http://sourceforge.net/projects/fuego/
For a list of major changes since Fuego 0.4.1, see
http://fuego.svn.sourceforge.net/viewvc/fuego/trunk/NEWS?revision=1394
Martin (for the Fuego contributors)
___
If someone created a semeai test suite competition for good scores would
drive progress .
I am sure every program has such a suite.
For example, Fuego's can be found at
http://fuego.svn.sourceforge.net/viewvc/fuego/trunk/regression/
Relevant tests are:
semeai_big.tst
uct_semeai.tst
So, let us have open-source program X and slightly modified version
it X'. What is the easiest way to run say 1000 super-fast games
between them? I hope there already exists some scripts or programs
to do this.
For testing Fuego in self-play and against other engines we use
CALL FOR PAPERS
===
Monte-Carlo Tree Search: Theory and Applications
ICAPS 2011 Workshop
http://icaps11.icaps-conference.org/workshops/mcts.html
Important Dates
==
Submission deadline: February 11, 2011
Author
I started a blog about computer Go on
http://computer-go-adventures.blogspot.com/
So far there are three posts on UEC cup (long), John Tromp's shodan bet games
against Many Faces (short) and the upcoming BIRS workshop (short). Enjoy!
Martin
How can i set up Fuego to do a fixed number of simulations/move?
For example:
uct_param_player ignore_clock 1
uct_param_player max_games 3000
And how many simulations do i need to make it play at around 1k level?
Which board size? On 19x19, I do not think it plays 1 kyu yet. Maybe with fast
Martin, how close were the 2nd and 3rd choices?
The exact numbers vary each time you run it. But all three moves kept getting a
substantial number of simulations throughout as I was watching it. For a more
precise answer, you'd have to run it maybe 100 times and trace the number of
At the exhibition game against Tei Meiko 9 Dan after the UEC cup, I resigned
for Fuego when the position was hopeless. However, Fuego's evaluation at that
time was still positive.
http://jsb.cs.uec.ac.jp/~igo/eng/result_ex/Fuego-MeikouTei.sgf
For example, at move 154 its value was 0.75 even
Dave,
I ran current Fuego on my laptop for 600K simulations. Similar to the other
programs, it likes simply C7 with a .82 evaluation. However, the bad moves M19
and N19 are second and third in number of simulations. So there seems to be
some systematic problem with the playouts and/or the tree
Is it possible to do lock free expansion with a transposition table as well?
-Magnus
Yes, Rémi has done this in Crazy Stone. His results are very impressive.
http://www.mail-archive.com/computer...@computer-go.org/msg07611.html
Martin___
I think there are three ways the tree is stored,
As a tree, with pointers
As a tree (with pointers), and with a separate transposition table - I think
Fuego does this
As a hash table only, with no pointers between nodes. This is what Many
Faces does.
Fuego only has a single
Yes. The theory behind this is described in Berlekamp's The Economist's View
of Combinatorial Games
http://www.msri.org/publications/books/Book29/files/ber.pdf
Martin
It is possible to calculate the importance of tedomari.
Suppose we play a game ignoring tedomari. As the value of
Hi Olivier,
do you think those are program-specific or a general problem with UCT (or more
likely the way simulations are run)?
Can you post a few specific cases?
In Fuego, we also have many performance test failures that do not seem to go
away, even at long time limits. But we never tried
32 matches
Mail list logo