break
> shenanigans. There is a defect in the script I use to convert
> https://www.gokgs.com/tournEntrants.jsp?sort=s=1068 to a crosstable.
> I'll probably leave the script as it is, and put this right with manual
> edits.
>
> I think Robert Waite reads this list, maybe he'll explain how
If you want something quick could use any engine that supports the gtp
command "final_score". It doesnt give win rate but will report who it
thinks won and by how many points. GnuGo.. pachi or commercial engines
likely all support it.. but calculation in early game can lag.. like 30s to
1min for
Estimating winrate is actually a very challenging problem.. particularly in
early and midgame. AlphaGo did create a value network that could help.. but
dont think any value networks are publicly available and even then cant
trust till later in game.
Could get guesses for next moves from a
out trying to see how they converged so quickly on GoGoD (or if the
implementation I am using is horribly flawed). Play strength seems to say
the network is working... so doubt the implementation is completely out of
whack.
On Wed, Aug 24, 2016 at 1:01 AM, Robert Waite <winstonwa...@gmail.com>
. humans. With just a network evaluation... that
is pretty dang impressive. I guess there might be weaknesses that humans
could figure vs. just the networks without search... but still... I'd be
pretty happy.
On Wed, Aug 24, 2016 at 12:30 AM, Robert Waite <winstonwa...@gmail.com>
wrote:
>
@Detlef It is comforting to hear that GoGoD data seemed to converge towards
51% in your testing. When I ran KGS data... it definitely converged more
quickly but I stopped them short. I think it all makes sense if figure 5 of
the DarkForest paper is the convergence of KGS data... and it doesn't
I had subscribed to this mailing list back with MoGo... and remember
probably arguing that the game of go wasn't going to be beat for years and
years. I am a little late to the game now but was curious if anyone here
has worked with supervised learning networks like in the AlphaGo paper.
I have
* - MoGo was using 5% of Huygens (instead of 25% against Kim);
* - there were some software improvements
* - MoGo won 2 out of 3 games in 9x9 (even games)
* - MoGo won with handicap 5 in 19x19 against the 6D player
That is interesting... it used 1/5th of the processing power and
got approximately
Just in case anyone hadn't seen the correction yet...
*
CORRECTION: *The EJ misquoted David Doshay in our 8/7 report on Computer
Beats Pro At U.S. Go Congress. What I said is that computer programs have
improved 7 to 9 stones in the last few years, [not We've improved nine
stones in just a year
Yes, but exhausitve search does not improve your player by 63% (eg.)
for a doubling in CPU time.
This part was done in an empirical scalability study. Please check the
archives of the list.
In the (inifinite) limit minimax+evaluation-function would find the
perfect move
too, but UCT/MC
play you probably meant it in a different sense than I read it
and many people on here might have understood what you were trying to point
out. I'm not trying to spray your parade with yellow rain... I just feel
that there are still many unknowns.
On Mon, Aug 11, 2008 at 12:23 PM, Robert Waite [EMAIL
You clearly don't understand the principles of alpha/beta pruning. It
is an admissible technique which means it guarantee's the same result
as searching the entire tree, but only requires a very tiny subset of
the entire tree.
Okay... congratulations... you are right... if you are able to
You clearly don't understand the principles of alpha/beta pruning. It
is an admissible technique which means it guarantee's the same result
as searching the entire tree, but only requires a very tiny subset of
the entire tree.
Okay... congratulations... you are right... if you are able to
Steve,
You mentioned three proofs relating to go... could you post the links to the
papers?
it makes no sense to ask if there is a mathematical proof
of anything related to humans.
I didn't ask for a mathematical proof saying if a computer can beat a human.
I asked in a roundabout way if this
* whether or not computers can beat humans at go on a
* 19x19 board in a reasonable amount of time is unrelated
* to mathematics.*
Because solving the game is not a prerequesite for beating the humans.
There are very obvious examples(chess)
I never questioned that. The way I read Steve's
at 2:54 PM, Robert Waite [EMAIL PROTECTED] wrote:
I'm curious what you guys think about the scalability of monte carlo with
UCT. Let's say we took a cluster like that which was used for the Mogo vs.
Kim game. Then lets say we made 128 of these clusters and connected them
together efficiently
to me.
On Sun, Aug 10, 2008 at 1:14 PM, Robert Waite [EMAIL PROTECTED]wrote:
* The MCTS technique appears to be extremely scalable. The theoretical
* * papers about it claim that it scales up to perfect play in theory.
** We agree here that this is not true of course.
*
No, I
Hmm.. I dunno.. I think there are a lot of ideas floating around but some
miscommunications.
So the aim is to devise a computer that will beat the strongest human
players of go.
I hear that Monte-Carlo with UCT is proven to be scalable to perfect play.
It seems that this is essentially saying...
I don't know how you can say that. The empirical evidence is
overwhelming that this is scalable in a practical way but more
importantly it's been PROVEN to be scalable. If you throw the word
practical in there then you are no longer talking the language of
mathematics, theory and proofs so
Well... I think I have hunches just as you do. And I think we both express
our hunches on here.
Diminishing returns is not really my theory.. I am just looking at
alternative ways of viewing the datapoints. Let's say you have two computers
and both of them focus only on solving local situations.
there are no problems that would take infinite time or infinite
space. there are problems that cannot be solved no matter
how much space or time you give a computer, but that's a
different matter altogether, and go isn't one of those problems.
How do you know what class go belongs in?
I was in the KGS room for a couple of hours before the match and a couple
after. I was very surprised by the result as many were.
There still is a lack of clear information about the event. For example,
when Kim said that the computer plays at maybe 2 or 3 dan... does he mean
professional or
Yeah.. the misclick question is another fuzzy point. There was a lot of
debate in the actual game about what was happening... but there is the
difficulty of having weak players and strong players commenting. The only
person who really knew what was happening and the direction of play is Mr.
Kim.
Oh yeah... I downloaded the final game from KGS and the sgf file seems to be
missing the small review that Mr. Kim gave at the end. He did not write
comments... he seemed to be doing it for those that were in the room. It
might be of interest to those that are interested in what he was thinking...
Yes... I do hope that more interest is sparked by this match. I had heard
that one of the big guys from Deep Blue now works for MS Research in Asia.
He had written a paper that I am sure most here have already read.. a title
similar to Cracking Go. I am sure he would be delighted by these results.
Well.. I disagree that too much significance is being made of it.
It is quite clearly a record. Handicap stones are a fundamental part of go.
It is uninteresting for human players to play an even game where one player
is incredibly stronger. There might be some recreational value.. but
I might come off as being strongly opinionated on the topic.. but I have
been of the opinion for a while that maybe playing go is a problem that
can't be solved by computers. I kinda want p != np and for us to be confined
by mathematics (sorry).The general taunt from my side is that A computer
can
well, in opposition to the p neq np problem, this is a fixed
boardsize. it's an engineering, optimization, and special-purpose
algorithm issue at this point. no need for any solution to work
for all boardsizes in some measurable, scalable way.
I don't necessarily think that go is
go is worse than np-complete, it's pspace-complete.
Well.. it would really depend on what you mean by solve go. If you mean to
solve it like they have with 5x5 for all possible moves... I don't know if
it is clear that 19x19 has the same properties. Ole Wikipedia, which very
well may be
* Besides... solving a
** pspace-complete problem would require infinite memory... isn't
that correct?
*
nope.
I flipped memory and time there. If pspace-complete is not in p, then it
will be a big problem trying to solve it without infinite time. That doesn't
seem like an ideal situation for
At worst we will just have to wait until robots take over the world in 20
years.
I would hope there wouldn't be a war... I'll join the robots. No need
for a body.
___
computer-go mailing list
computer-go@computer-go.org
31 matches
Mail list logo
