Just purely curiosity: How strong is Leela now? googling up gives that it
is better than best humasn already? Is that true?
Petri
ma 28. tammik. 2019 klo 23.31 "Ingo Althöfer" (3-hirn-ver...@gmx.de)
kirjoitti:
> Hello,
>
> a central quote from the Leela Github blog at
>
while working at Nokia we revieved what innovations were worth patenting
and how to deal with those we did not see important enough. Sometime we
paid a reseacher from acemia to write a paper and submit for some
conference. There were other methods. But filing just for prior art is way
too
elo-range in 9x9 smaller than 19x19. One just cannot be hugelyl better than
the other is such limitted game
2018-02-23 21:15 GMT+02:00 Hiroshi Yamashita :
> Hi,
>
> Top 19x19 program reaches 4200 BayesElo on CGOS. But 3100 in 9x9.
> Maybe it is because people don't have much
Seems suprisingly strong. Given that no super vcluster availab,´le for
trainning. Have at least on accoutn rated would be nice since in unrated
games people experiment quite a lot at cost of playing well.
2017-12-29 22:02 GMT+02:00 Brian Sheppard via Computer-go <
computer-go@computer-go.org>:
>But again: For instance, when a eight year old child starts
>to play violin, is it helpful or not when it had played
>say a trumpet before?
It would be and this is well known in practice. Logic around the music is
the same so hw would learn faster. In the very long run there might be no
wanted
So If I load latest leela into a laptop (being about KGS 4k) I would expect
to demolished even on 9 stone handicap. Nice
Petri
2017-11-15 13:55 GMT+02:00 Darren Cook :
> > Zero was reportedly very strong with 4 TPU. If we say 1 TPU = 1 GTX 1080
> > Ti...
>
> 4 TPU is 180
and we can allways come up with bizarre situation like casualties insidet
the vehicle vs casualties to persoons outside the vehicle. I am pretty sure
this will a long discussion with huge research gaps on ethics as well as
in engineering
2017-10-31 7:00 GMT+02:00 Robert Jasiek :
intuition is handy word for truly automated information processing i.e
subconscious. And everything that train conscious decission making trains
also the subconscious/intuiton. Intuiton nothing mythical just automation
achieved via training
2017-10-29 5:08 GMT+02:00 Thomas Rohde
wn
> intuition fairly well.
>
> Do we want to sit down and analyze the best human player's intuition?
> Perhaps. But certainly not to improve the best computer player. It can
> already crush all humans at pretty much every strength.
>
> s.
>
>
> On Fri, Oct 27, 2017 at 1
way more important the result. Obviousl I canno tprove
my point as my evidence is anecdotal
PP
2017-10-26 17:54 GMT+03:00 Robert Jasiek <jas...@snafu.de>:
> On 26.10.2017 08:52, Petri Pitkanen wrote:
>
>> Unfortunately there is no proof that you principles work better than thos
Unfortunately there is no proof that you principles work better than those
form eighties. Nor there is any agreement that your pronciples form any
improvement over the old ones. Yes you are a far better player than me and
shows that you are
- way better at reading
- have hugely better go
O'Flaherty <jim.oflaherty...@gmail.com>:
> Couldn't they be useful as part of a set of training data for newly
> trained engines and networks?
>
> On Oct 23, 2017 2:34 AM, "Petri Pitkanen" <petri.t.pitka...@gmail.com>
> wrote:
>
>> They are free to use
They are free to use in any attribution. Game score is a reflection of
historical fact and hence not copyrightable. Dunno what use them are to
anyone though.
Petri
2017-10-23 2:29 GMT+03:00 Lucas Baker :
> Hi Robert,
>
> The AlphaGo Zero games are free to use with proper
1) There is no such thing and I do doubt if it ever will exist. Even humans
fail elaborate why they know certain things
2) If we are talking about new one. Very few people seen it playing so I
guess we lack the data. For the old we know it made errors, dunno if
analysis points why. Neural nets
Cost reduction in IC has reached or is reaching its limits. Intels 5n techk
is not really a 5n and 5n is not really reachable. Not at least without
some seriously new physics and even then there will be hard limits like
quantum un--certainty. This particular chip may get cheaper if it is ever
done
Has to be truly far. Even in chess best estimates are that current
computers are still few hundred ELO points away from perfect. In chess on
can make estimates based what draw rates current best ones would get
against perfect play. Loads of guess work but still reasonable. In go
difference must
never heard of one for GO. High School kids in Finland made one for chess.
And to get proper picture camera had to located avbovr the board.I suspect
it has to same way in go. So would require some kind on phone stand
Here is start of such project:
https://gogamerecorder.wordpress.com/
SO I
he is muted for a good reason. And "classical " approach was done for
decades did no lead much to anything. So I would be interested on the paper
ONLY if author has implemented his/her ideas and measured performance.
Otherwise it would be waste of time with at least 95% probability
Petri
Would the expected improvement be reduced training time or improved
accuracy?
2016-06-11 23:06 GMT+03:00 Stefan Kaitschick :
> If I understood it right, the playout NN in AlphaGo was created by using
> the same training set as the one used for the large NN that is
Chess was popular everywhere so the barriers were relatively small. As one
chess writer said it. There are moer chess titles written than all other
hobby titles combined. Dunno who reads all of them.
But I do doubt if strong go programs give too much for analysis. Even if
they are 1p and can show
There are several open source go programs. I would start by investigating
Fuego and Pachi code
2016-04-10 11:34 GMT+03:00 Gonçalo Mendes Ferreira :
> There isn't a lot of info on this[1], so it will probably be a hard
> journey for a fast representation. But the things a Go board
Since there are only two possible outcomes it pretty much normal. Actually
binomial which will converge to normal given enough samples
Only thing that cans distort is that consecutive games are not
independent (which
is probably the case but do they have positive or negative correlation?)
I would second this. computers in chess do not teach anything. Computer can
show you the great move but cannot explain it.
Probably as hard problem to crack as was making a good computer go
2016-03-14 16:22 GMT+02:00 Robert Jasiek :
> On 14.03.2016 08:59, Jim O'Flaherty wrote:
interpretation
petri
2016-03-14 16:11 GMT+02:00 Robert Jasiek <jas...@snafu.de>:
> On 14.03.2016 09:33, Petri Pitkanen wrote:
>
>> And being 600 elo points above best human you are pretty close to best
>> possible play.
>>
>
> You do not have any evidence for
l
> enjoy the utter beauty of its producing something human minds could not
> directly perceive much less achieve. I am sure there are others. These were
> the ones that just sprung to mind without much effort.
>
>
>
> On Mon, Mar 14, 2016 at 2:07 AM, Petri Pitkanen <
> pe
Even though the chess SW analyzes less positions/second than earlier it
does not mean it is less dependent of good HW. Complex move selection and
smart evaluation do need more CPU as well. There are advances like NULL
move pruning that reduce the amount of CPU required but still it is very
much HW
This time I think game was tougher. Though too weak to judge. At the end
sacrifice a fistfull stones does puzzle me, but again way too weak to
analyze it.
It seem Lee Sedol is lucky if he wins a game
2016-03-10 12:39 GMT+02:00 Petr Baudis :
> On Wed, Mar 09, 2016 at 09:05:48PM
Opent to intepretation if this method is brute force. I think it i. Uses
huge amounts of CPU power to run simulations and evaluate NN's. Even in
chess it was not just about tree search, it needs evaluationfunction ot
make sense of the search
2016-02-24 6:52 GMT+02:00 muupan :
>
Still no time limits. I would assume that very short time limits help
computer, not very sure though
2016-02-06 14:40 GMT+02:00 "Ingo Althöfer" <3-hirn-ver...@gmx.de>:
> Hello,
>
> on the Alpha-Go website a date for the match between
> Lee Sedol and Alpha-Go is given:
> 5 rounds, to be played
Welll, David is making a product. Making a product is 'trooper' solution
unless you are making very specific product to a very narrow target group,
willing to pay thousands for single license
Petri
2016-02-04 23:50 GMT+02:00 uurtamo . :
> David,
>
> You're a trooper for doing
At least on digital filter time increases non-linearly - you can think NN
as non-linear FIR. And multilayer structure should make this harder, if you
think of it . So some tricks to speed it up might be necessary. dunno about
NN but on digital filters one trick was to train first part of filter
Explaining why the move is good in human terms is useless goal. Good chess
programs cannot do it nor it is meaningful. As the humans and computers
have vastly different approach to selecting a move then by the definition
have reasons for moves. As an example your second item 'long-term aji', For
i think similar approaches have been done. I can recall seeing it. Though
in Backgammon they did train only by endresult and seemed to work fine.
Originally anyway, now the have separate NN-for certain phases of the game
2016-01-30 18:07 GMT+02:00 Xavier Combelle :
> I
I do not think such exercise would give any meaningful results. NN would
not imitate it's 'hero' 1-1 not even close
Funny such discussion keep on going on in Chess and Go, in chess i think
Steiniz would be wiped of board by best players of today, but still he
would be way better as he created
I think such analysis might not bee too usefull. At least chess players
think it is not very usefull. Usually for learning you need "wake-up" your
brains so computer analysis without reasons probabaly on marginally useful.
But very entertaining
2016-01-28 13:27 GMT+02:00 Michael Markefka
*"Seki means a constellation on the go board with two*
*living neighboring groups: one by Black, the other oneby White. Each of
the groups has only one eye"*
Why would you need an eye for seki?
http://senseis.xmp.net/?Seki
Shared liberties is good enough and quite typical in my limited
Yes scraping for large amounts of data from a smallish server is not
really polite. May overload the server. Besides quite inefficient. You
could make a request to owner of site instead. Assuming you can present
good enough reason you might get lucky
2015-11-13 8:39 GMT+02:00 Josef Moudrik
r weak players Go than in chess, because doing a Null move is far easier
2015-11-05 13:39 GMT+02:00 Petr Baudis <pa...@ucw.cz>:
> On Thu, Nov 05, 2015 at 09:03:38AM +0200, Petri Pitkanen wrote:
> > 2015-11-05 0:04 GMT+02:00 Hideki Kato <hideki_ka...@ybb.ne.jp>:
> >
&
to be collected by the site
operator. Webspider could overload the system and no interface exist that
would be usefull for collecting the data
2015-11-05 18:26 GMT+02:00 Michael Alford <m...@aracnet.com>:
> On 11/5/15 7:19 AM, Petr Baudis wrote:
>
> On Thu, Nov 05, 2015 at 02:42:20PM +0200,
and in Go one move advantage need that your 1st pro-level mode works
together with your subsequent non-pro-moves
2015-11-05 14:55 GMT+02:00 Christoph Birk :
>
> On Nov 5, 2015, at 4:44 AM, Nick Wedd wrote:
> > However, there's a powerful
Let alone we do not have even sufficient understanding of perfect play to
say what is correct komi in absolute sense. Nor it is it even meaningful
concept. Correct komi is a komi that produces about 50/50 result. Obviously
komi that will result in 50/50 for professionals will probably favour white
I think very few people here do not know message passing style of
programming. I just not suited problem at hand. Not very cPU efficient.
This is high speed simulation anyways
2015-10-02 16:53 GMT+03:00 djhbrown . :
> .
> "sharing code is typically not going to be
David said "estimate final score" which implies that all relevant things
are factored in, merely the unit of estimation is territory. Just like in
chess there are several things factored in - other than material - and all
are estimated as pawns.
I guess expert systems really are a dead end in
I played few games against bots in arimaa.com and they seemed to react. I
think the eval can be used for that?
I did not find the game interesting. Just being hard for computers does not
make it fun. So I quit playing after few games
Petri
2015-04-23 18:32 GMT+03:00 Stefan Kaitschick
Assuming you are using some sensible OS there better ways profile than
sample like oprofile for linux. There is similar thing for FreeBSD I
think. No instrumentation san sampling gets automated
Petri
2015-03-30 8:05 GMT+03:00 hughperkins2 hughperki...@gmail.com:
40% sounds pretty high. Are
They also all lose games on endgame same manner. Having won a game by 30 pts
they start giving away those points for - sometimes - imaginary safety,
allowing other player to come within striking distance. Some sort dynamic
komi would be nice in endgame, but would probably not work.
In Handicaps
2009/12/1 Rémi Coulom remi.cou...@univ-lille3.fr
David Fotland wrote:
I'd like to thank all of the people who organized the UEC tournament for
providing machines and operators to allow Many Faces and others to
participate. I'd like to suggest that the UEC organizers consider using a
Swiss
Well No, this games game lot harder. Even when point matter, 1st goal is to
win the game in traditional sense to get any points at all. Which make just
as hard as normal game. Then comes huge risk assesment risks involved. Lets
assume - not so rare case - that you can go for the throat or attack
Only papers I can recall are from seventies (assuming you mean academic
papers) from Wilcoxx. I may have electrical copies. Not sure though. I
managed to find some of them from ACM site.
That paper described position based approach where each and every stage was
stored into datastructure, kinda
On anecdotal evidence:
Manyfaces on medium time settings KGS = 2k (accounts manyfaces and
manyfaces2)
Manyfaces1 playing round 10 sec/move is able maintain 1d rank.
So by reducing oppponents thinking time bot gets relative advantage of
3stones.
Also in chess it is uusually considered that
I cant recall any offocoal challenges. I do remember some such statement in
some other challenge, but failed to google it up.
Human computer chess challenges are not likely to happen anymore. What would
be the point for human? Hydra could probably beat anyone. And as processors
get faster any of
Hello,
Are there more peculiar situation that will cause problems for MCTS apart
from the three I know.
1. Nakade (this is partuially solved in most of the programs)
2. Semeais
3. Double Ko.
Last one was new to me. See
http://files.gokgs.com/games/2009/10/26/ManyFaces-Hyoga.sgf
In that game
That is well known fact of go, that usually defence is easier. But evidence
is anecdotal. Getting real evidence from real games cannot be automated as
all concept involved are rather vague and difficult to classify. Hence I am
willing to accept such information as passed on to me in books like
I dont have but harvesting those from KSG archives should not be too
difficult using Pelr/mechanize or similar system. Remembering to put sleeps
into scripts as not to overload the system. Start f.ex by getting the list
of 100 player and see through their games.
Petri
2009/10/20 Petr Baudis
Not really a compuetr Go issue, but I do not think that great wall is
superior even when completed. It is not too bad but it needs a definite
strategy from wall owner. I.e building side moyos using wall as a roof and
hoping that the other guy gets nervous and jumps in. So by being patient is
Neural network tend to work well in those cases where evaluation function is
smooth, like backgammon. Even inbackgammon neural networks do give good
results if situation has possibility of sudden equity changes like deep
backgames and deep anchor games. Top backgammon programs 3-ply search on top
Maybe they are long way from giving handicaps to you. But best of bots
in KGS are around 2k and there are hundreds of 9k and weaker players
present there at all times. So being able to play white is worthy
thing at least for commercial bot.
Petri
2009/8/13 Christoph Birk b...@ociw.edu:
On Aug
2009/4/23 terry mcintyre terrymcint...@yahoo.com:
Programs which get semeai and seki right every time might be a few stones
stronger. They'd certainly be more valuable as teaching tools. In the game
above, a stronger program would have exploited my earlier weakness; this
would have encouraged
Because your time measurement has gone wrong. You get 0 seconds in
time hence kpssa in infinity.
Petri
2009/4/23 Michael Williams michaelwilliam...@gmail.com:
Here is my full set of numbers. I wonder why the known kpps/GHz but unknown
kpps.
= Benchmarking, please wait ...
= 20
2009/4/8 Zhiheng Zheng zhiheng.zh...@gmail.com:
I think most of test are designed by people who is stronger than best
computer go program. So if MC program fail to pass a test, it is most likely
MC is wrong. MC program is strong in some aspect, but week in other aspect.
And the test suit is
Libego has one type of mercy rule.
Petri
2009/2/27 Seth Pellegrino se...@lclark.edu:
Hello list,
I've managed to track the idea of a mercy rule in monte-carlo playouts back
to a mail sent to this list by David Hillis:
http://computer-go.org/pipermail/computer-go/2006-December/007478.html
Commercial market for Go software is in Japan in Korea. Western player
do not make significant numbers and Chinese probably find bettre uses
for money - although there more reach Chinese people than people in
Finland.
Petri
2008/11/21 Michael Gherrity [EMAIL PROTECTED]:
Hi,
I have read that
2008/9/4 Rémi Coulom [EMAIL PROTECTED]:
only 5k, so I cannot really tell. But when I see the horrors it plays in
some games, I suppose it must play much stronger than 1k in some other games
in order to get a rating of 1k.
Look for instance at these two games:
a win:
2008/7/28 David Fotland [EMAIL PROTECTED]:
The traditional programs are around 10 kyu, but the new ones are 2 to 4 kyu,
at least on KGS. I've seen some handicap games against dan players that are
consistent with these ratings.
It wouldn't surprise me to see 1 dan from an MC program before
Is the KGS manyfacesofgo MC version or traditional. Just seems to
tenuki quite MC fashion
Petri Pitkänen
--
Petri Pitkänen
e-mail: [EMAIL PROTECTED]
___
computer-go mailing list
computer-go@computer-go.org
2008/4/9, Don Dailey [EMAIL PROTECTED]:
Since I sell software, building Linux apps is out of the question, since
Linux users will insist that I give them my work for free.
I don't have any issue whatsoever with making money by selling software
either. I'm not one of those guys
Is just my problem or do others have it. When I try to download for example:
http://www.lri.fr/~teytaud/SGF/2008/01/30/17067.sgf
I Get 403 response with following line:
You don't have permission to access /~teytaud/SGF/2008/01/30/17067.sgf
on this server.
Cheers,
Petri
--
Petri Pitkänen
e-mail:
2008/1/30, Don Dailey [EMAIL PROTECTED]:
It would get it eventually, which means this doesn't inhibit scalability.
Having said that, I am interested in this. Is there something that
totally prevents the program from EVER seeing the best move?I don't
mean something that takes a long
Hello,
Has anyone experimented with libego parameters? What would be
reasonable starting point on configurations on 19x19 board? I am
considering using GnuGo to give the moves MC is allowed to simulate
for first couple of plies in simulation and then pure random. Just to
see if it makes any sense
2008/1/22, Eric Boesch [EMAIL PROTECTED]:
On Jan 22, 2008 1:43 AM, Petri Pitkanen [EMAIL PROTECTED] wrote:
Even top MC programs fail to see that a group with 3 liberties with
no eyes is dead.
A 3-liberty group with no eyes has a 100% chance to die during
playouts unless a surrounding
2008/1/22, Erik van der Werf [EMAIL PROTECTED]:
In the future, when humans are consistently defeated by computers on
19x19 and the remaining players move up to a more 'interesting' size,
will you be claiming that 19x19 isn't Go either?
E.
Maybe I will, but 17x17 is quite like 19x19, While
So far I played these MC programs at it seems they are doing well
against humans mostly because the moves they play are bizarre and some
times throw unreasonable contact fight challenges. They win more often
than they deserve just because many weak players (like KGS 4k level
players are) quite
2007/12/11, terry mcintyre [EMAIL PROTECTED]:
With Go, there are many situations which can be read out precisely, provided
that one has the proper tools - ladders, the ability to distinguish between
one and two eyes; the ability to reduce eyespaces to a single eye with an
appropriate
2007/12/11, Don Dailey [EMAIL PROTECTED]:
Hi Petri,
I happen to think that MC is the most human like approach currently
being tried
Ye in sense Alpha-Beta is human like. It one feature we do and takes
it to extreme. And using different method of evaluation.
.
The reason I say that is that
There is something that the latest Monte Carlo programs have in common
with the best chess programs - and seems to be the right way to
structure a game tree search.Your selectivity should be
progressive. In order to do this correctly you must re-visit nodes
many times. Chess programs
2007/2/16, Nick Apperson [EMAIL PROTECTED]:
trouble. Also, the alternative is usually function pointers which have
atleast 50 times the overhead of a function object. Correct me if I'm
wrong.
- Nick
function objects really cannot be 50 times more efficient as function
pointer are rather
2007/1/11, Don Dailey [EMAIL PROTECTED]:
50 X speedup sound rather impressive but it's not that much. It's
probably
made go programs about 2 or 3 stones stronger over the few years that it
took to get hardward 50X faster about what you would expect.
But it is hardly that much. Current
opponent and eventually could have passed for free. Had game been under
Japanese rules I would have been 'forced' to think whether reply was
needed and thus think a lot longer time for replies and possibly lost on
time because reply would have been needed probably too often.
Conclusion: Under
2007/1/4, Don Dailey [EMAIL PROTECTED]:
No, this inhibits the application of skill. A silly invasion that
wastes time is punished in all rules sets, but in Chinese it may not
be silly if it doesn't waste time - Japanese rules unfairly defines
these moves as silly.
It is silly if opponents
Or use p2p and the pirate bay. Using serch word SGF you should find
about 40 000 game collection from moyo-go.
Or even easier The Torrent:
http://torrents.thepiratebay.org/hashtorrent/3420315.torrent/40_683_Professional_Go_Games_Collection.3420315.TPB.torrent
As game records are not
80 matches
Mail list logo