Re: [Computer-go] CGOS source on github

[uurtamo]

also frankly not a problem for a rating system to handle.

a rating system shouldn't be tweaked to handle eccentricities of its
players other than the general assumptions of how a game's result is
determined (like, does it allow for "win" and "draw" and "undetermined" or
just "win").

s.


On Fri, Jan 22, 2021 at 6:29 AM David Wu  wrote:

> On Fri, Jan 22, 2021 at 8:08 AM Rémi Coulom  wrote:
>
>> You are right that non-determinism and bot blind spots are a source of
>> problems with Elo ratings. I add randomness to the openings, but it is
>> still difficult to avoid repeating some patterns. I have just noticed that
>> the two wins of CrazyStone-81-15po against LZ_286_e6e2_p400 were caused by
>> very similar ladders in the opening:
>> http://www.yss-aya.com/cgos/viewer.cgi?19x19/SGF/2021/01/21/73.sgf
>> http://www.yss-aya.com/cgos/viewer.cgi?19x19/SGF/2021/01/21/733301.sgf
>> Such a huge blind spot in such a strong engine is likely to cause rating
>> compression.
>> Rémi
>>
>
> I agree, ladders are definitely the other most noticeable way that Elo
> model assumptions may be broken, since pure-zero bots have a hard time with
> them, and can easily cause difference(A,B) + difference(B,C) to be very
> inconsistent with difference(A,C). If some of A,B,C always handle ladders
> very well and some are blind to them, then you are right that probably no
> amount of opening randomization can smooth it out.
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] CGOS source on github

[uurtamo]

It's a relative ranking versus who you actually get to play against.

Sparsity of actual skill will lead to that kind of clumping.

The only way that a rating could meaningfully climb by playing gnugo or
your direct peers is going to happen exponentially slowly -- you'd need to
lose to gnugo twice less often (or win all the time over twice as many
games) to get more points. So although it would eventually increase, it
would flatten out pretty quickly.

Good point about mcmc. A more dramatic approach would be to remove gnugo
altogether.


On Mon, Jan 18, 2021, 6:41 AM Rémi Coulom  wrote:

> Hi,
>
> Thanks to you for taking care of CGOS.
>
> I have just connected CrazyStone-57-TiV. It is not identical, but should
> be similar to the old CrazyStone-18.04. CrazyStone-18.04 was the last
> version of my program that used tensorflow. CrazyStone-57 is the first
> neural network that did not use tensorflow, running with my current code.
> So it should be stronger than CrazyStone-18.04, and I expect it will get a
> much lower rating.
> A possible explanation for the rating drift may be that most of the old MC
> programs have disappeared. They won easily against GNU Go, and were easily
> beaten by the CNN programs. The Elo statistical model is wrong when
> different kind of programs play against each other. When the CNN program
> had to get a rating by playing directly against GNU Go, they did not manage
> to climb as high as when they had the MC programs between them and GNU Go.
> I'll try to investigate this hypothesis more with the data.
>
> Rémi
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] 30% faster with a batch size of 63 instead of 64!

[uurtamo .]

Nice job! And the graph makes it super clear how the edge effects work.

s.

On Sat, May 9, 2020, 2:19 PM Rémi Coulom  wrote:

> Hi,
>
> I am probably not the only one who made this mistake: it is usually very
> bad to use a power of 2 for the batch size!
>
> Relevant documentation by NVIDIA:
>
> https://docs.nvidia.com/deeplearning/performance/dl-performance-convolutional/index.html#quant-effects
>
> The documentation is not extremely clear, so I figured out the formula:
> N=int((n*(1<<14)*SM)/(H*W*C))
>
> SM is the number of multiprocessors (80 for V100 or Titan V, 68 for RTX
> 2080 Ti).
> n is an integer (usually n=1 is slightly worse than n>1).
>
> So the efficient batch size is 63 for 9x9 Go on a V100 with 256-channel
> layers. 53 on the RTX 2080 Ti.
>
> There is my tweet with an empirical plot:
> https://twitter.com/Remi_Coulom/status/1259188988646129665
>
> I created a new CGOS account to play with this improvement. Probably not a
> huge different in strength, but it is good to get such an improvement so
> easily.
>
> Rémi
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Crazy Stone is playing on CGOS 9x9

[uurtamo .]

And this has no book, right? So it should be badly abused by a very good
book?

s.

On Fri, May 8, 2020, 3:28 PM David Wu  wrote:

> I'm running a new account of KataGo that is set to bias towards aggressive
> or difficult moves now (the same way it does in 19x19 handicap games), to
> see what the effect is. Although, it seems like some people have stopped
> running their bots.  Still maybe it will be interesting for the
> remaining players, or any others who decide to re-turn-on their bot for a
> little while. :)
>
> It seems like some fraction of the time, it now opens on 5-5 as black,
> which is judged as worse than 4-4 in an even game, but presumably is more
> difficult to play. I suspect it will now start to lose a noticeable number
> of games now due to overplaying, and there's a good chance it does much
> worse overall. Even so, I'm curious what will happen, and what the draw
> rate will be. Suddenly having some 5-5 openings should certainly add some
> variety to the games.
>
> On Thu, May 7, 2020 at 12:41 PM David Wu  wrote:
>
>> Having it matter which of the stones you capture there is fascinating.
>> Thanks for the analysis - and thanks for "organizing" this 9x9 testing
>> party. :)
>>
>> On Thu, May 7, 2020 at 12:06 PM Rémi Coulom 
>> wrote:
>>
>>> If White recaptures the Ko, then Black can play at White's 56, capture
>>> the stone, and win by 2 points.
>>>
>>> On Thu, May 7, 2020 at 5:02 PM Shawn Ligocki  wrote:
>>>
 Thanks for sharing the games, Rémi!

 On Thu, May 7, 2020 at 6:27 AM Rémi Coulom 
 wrote:

> In this game, Crazy Stone won using a typical Monte Carlo trick:
> http://www.yss-aya.com/cgos/viewer.cgi?9x9/SGF/2020/05/07/997390.sgf
> On move 27, it sacrificed a stone. According to Crazy Stone, the game
> would have been a draw had Aya just re-captured it. But Aya took the bait
> and captured the other stone. Crazy Stone's evaluation became instantly
> winning after this, the sacrificed stone serving as a threat for the
> winning ko fight, 18 moves later.
>

 Wow, I did not imagine how that move would be useful later! But the
 very end is confusing to my human brain, couldn't White move 56 retake the
 ko and win it? It seems like Black only has one real ko threat left (J4
 maybe). But White also has one huge threat left (D3), so it seems like
 White should win this ko and then be about 4 ahead with komi. Am I
 missing something?

 -Shawn
 ___
 Computer-go mailing list
 Computer-go@computer-go.org
 http://computer-go.org/mailman/listinfo/computer-go

>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Crazy Stone is playing on CGOS 9x9

[uurtamo .]

great book interface, by the way.

s.


On Wed, May 6, 2020 at 11:01 AM Rémi Coulom  wrote:

> Hi,
>
> I trained a neural network for 9x9, and it is playing on CGOS.
>
> The network has 40 layers (20 residual blocks) of 256 units. It is running
> on a Titan V GPU, with a batch of 64, at about 9k playouts per second.
>
> It is using an opening book that you can browse online there:
> https://www.crazy-sensei.com/book/go_9x9/
> Each node of the book was searched with 400k playouts.
>
> I will let it play for a few days.
>
> Rémi
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] codecentric Freestyle League - Great Final

[uurtamo]

BTW: there's a fairly straightforward way to evaluate the skill level of
the games on the whole. Is there any interest in that, or just the results?

steve

On Fri, Apr 12, 2019, 1:38 PM "Ingo Althöfer" <3-hirn-ver...@gmx.de> wrote:

> Hello,
>
> the final round of the codecentric Freestyle League is scheduled
> for Wednesday, April 24.
> https://www.althofer.de/codecentric-freistil.html
>
> However, the pairing that decides about the title, is
> Karlsruhe - Paderborn,
> and will start already on Sunday, April 21, 18:00 Central
> European Summer Time.
>
> Place is "Deutsche Ecke" on KGS.
>
> Spectators (of any shade between White and Black) are welcome.
>
> Ingo.
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] 0.5-point wins in Go vs Extremely slow LeelaChessZero wins

[uurtamo]

The human passenger is going to be asleep in the car and hopefully not
awakened by something as trivial as braking.

I recently understood how komi is being dealt with by Leela zero (or at
least by petgo). It's so kludgey and yet is 7d at kgs.

So let's just relax on the small optimizations. The big stuff is so
powerful that it's not even funny.

s.

On Tue, Mar 5, 2019, 2:11 PM "Ingo Althöfer" <3-hirn-ver...@gmx.de> wrote:

> Hi,
>
> recently, Leela-Chess-Zero has become very strong, playing
> on the same level with Stockfish-10. Many of the test players
> are puzzled, however, by the "phenomenon" that Lc0 tends to
> need many many moves to transform an overwhelming advantage
> into a mate.
>
> Just today a new German tester reported a case and described
> it by the sentence "da wird der Hund in der Pfanne verrückt"
> ("now the dog is going crazy in the pan", to translate it word
> by word). He had seen an endgame: Stockfish with naked king,
> and LeelaZero with king, queen and two rooks. Leela first
> sacrificed the queen, then one of the rooks, and only then
> started to go for a "normal" mate with the last remaining rook
> (+ king). The guy (Florian Wieting) asked for an explanation.
>
> http://forum.computerschach.de/cgi-bin/mwf/topic_show.pl?tid=10262
>
> I think there is a very straightforward one: What Leela-Chess-Zero
> with its MCTS-based searc) performs is comparable to the
> path all MCTS Go bots took for many years when playing winning
> positions against human opponents: the advantage was reduced
> step by step, and in the end the bot gained a win by 0.5 points.
> Later, in the tournament table, that was not a problem, because
> a win is a win :-)
>
> Similarly in chess: overwhelming advantage is reduced by lazy play
> to some small margin advantage (against a straightforward alpha-beta
> opponent), and then the MCTS chess bot (= Leela Zero in this case)
> starts playing concentratedly.
>
> Another guy asked how DeepMind had worked around this problem
> with their AlphaZero. I am rather convinced: They also had this
> problem. Likely, they kept the most serious examples undisclosed,
> and furthermore set the margins for resignation rather narrow (for
> instance something like evaluation +-6 by Stockfish for three move
> pairs) to avoid nearly endless endgames.
>
> Ingo.
>
> PS: thinking of a future with automatic cars in public traffic. The
> 0.5-point wins or the related behaviour in MCTS-based chess would mean
> that an automatic car would brake only in the very last moment
> knowing that it will be sufficient to stop 20 centimeters next to the
> back-bumpers of the car ahead. Of course, a human passenger would
> not like to experience such situations too often.
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

[Computer-go] ...

[uurtamo]

re: that leela game i posted, it's clearly buggy behavior -- filling in one
of its two last eyes, for instance.

s.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

[Computer-go] ...

[uurtamo]

funny leela behavior this morning in a selfplay game (truncated) in case
you'd like to see how things can go from time to time (this is with
05d10f27).  i thought it might have wider interest because it was so long
of a game and because it resulted in such a lopsided score:

641 (B M5) 642 (W A6) 643 (B E10) 644 (W G15) 645 (B E1) 646 (W A4) 647 (B
T6) 648 (W F1) 649 (B H4) 650 (W H5) 651 (B T3) 652 (W E1) 653 (B K13) 654
(W G1) 655 (B S4) 656 (W T7) 657 (B P8) 658 (W K4) 659 (B C2) 660 (W J14)
661 (B G14) 662 (W C7) 663 (B L13) 664 (W T6) 665 (B P12) 666 (W pass) 667
(B P3) 668 (W Q12) 669 (B Q19) 670 (W O1) 671 (B F19) 672 (W A8) 673 (B
Q11) 674 (W A7) 675 (B G15) 676 (W P1) 677 (B N5) 678 (W P2) 679 (B H12)
680 (W R4) 681 (B R16) 682 (W F2) 683 (B P11) 684 (W B7) 685 (B S5) 686 (W
R6) 687 (B S8) 688 (W S6) 689 (B D8) 690 (W Q5) 691 (B H6) 692 (W T5) 693
(B R4) 694 (W R5) 695 (B R11) 696 (W T7) 697 (B Q12) 698 (W T6) 699 (B H5)
700 (W pass) 701 (B M7) 702 (W L7) 703 (B pass) 704 (W pass) Game has ended.

Score: B+227.5

Winner: black

Uploading game: 5977c5f737a8405fa8f58b737e2e9ea2.sgf for network
05d10f27e549f3d1e34bc449d406fba18014f262d395cd67cec957829ef399df

Game data 04cf04f7f7bdcbc3980e769080c28b5742a66302125ab6c0b0090ced1e54a2f7
stored in database


s.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] GCP passing on the staff ...

[uurtamo]

That's a nice way to describe it.

What I've seen is that mid-fight it will just play elsewhere. Like hard mid
fight. Middle of the fight just walk away

On Wed, Jan 30, 2019, 2:25 PM Igor Polyakov  I saw a game where it decided connecting its group wasn't important since
> it could still make the cut off side sabaki later
>
> Approaches corner, decides it's ahead, lives later
>
> It doesn't care about keeping groups strong as much
>
> On Wed, Jan 30, 2019, 13:24 uurtamo 
>> They will abandon a fight to take bigger sente. It's super scary to watch.
>>
>> s.
>>
>> On Tue, Jan 29, 2019, 8:06 PM Robert Jasiek >
>>> On 29.01.2019 18:53, uurtamo wrote:
>>> > it's [...] about an insane need to keep sente. my only
>>> > takeaway other than reading out fights way way way in advance.
>>>
>>> I can confirm the necessity for keeping sente with respect for the
>>> endgame but would not be surprised it to also apply during opening and
>>> middle game. One of the greatest weaknesses of my pupils in the kyus is
>>> not to play all their sentes (other than privileges preserved for ko
>>> threats or liberties) before gotes. From my study, research of and book
>>> writing on the endgame during the previous 2.5 years, I have realised
>>> the importance of distinguishing gote from sente even if their
>>> difference is only a fraction of a point (but it can be up to ca. 5
>>> points per local decision) and of exceptionally playing gote instead of
>>> sente or vice versa depending on the global context. Every small mistake
>>> in evaluation about playing too long locally etc. amounts to a large
>>> total amount when all mistakes accumulate. Programs would notice such
>>> implicitly due to their smaller winning chances when making too many
>>> such mistakes.
>>>
>>> Reading out fights in advance very deeply I have only noticed a few
>>> times during programs' play but, of course, you are right. I simply have
>>> not studied their deep reading carefully enough to witness more
>>> incidents.
>>>
>>> --
>>> robert jasiek
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] GCP passing on the staff ...

[uurtamo]

They will abandon a fight to take bigger sente. It's super scary to watch.

s.

On Tue, Jan 29, 2019, 8:06 PM Robert Jasiek  On 29.01.2019 18:53, uurtamo wrote:
> > it's [...] about an insane need to keep sente. my only
> > takeaway other than reading out fights way way way in advance.
>
> I can confirm the necessity for keeping sente with respect for the
> endgame but would not be surprised it to also apply during opening and
> middle game. One of the greatest weaknesses of my pupils in the kyus is
> not to play all their sentes (other than privileges preserved for ko
> threats or liberties) before gotes. From my study, research of and book
> writing on the endgame during the previous 2.5 years, I have realised
> the importance of distinguishing gote from sente even if their
> difference is only a fraction of a point (but it can be up to ca. 5
> points per local decision) and of exceptionally playing gote instead of
> sente or vice versa depending on the global context. Every small mistake
> in evaluation about playing too long locally etc. amounts to a large
> total amount when all mistakes accumulate. Programs would notice such
> implicitly due to their smaller winning chances when making too many
> such mistakes.
>
> Reading out fights in advance very deeply I have only noticed a few
> times during programs' play but, of course, you are right. I simply have
> not studied their deep reading carefully enough to witness more incidents.
>
> --
> robert jasiek
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] GCP passing on the staff ...

[uurtamo]

so aside from the rumors, i've watched games that are simply mind-blowing.

it's, from what i've seen, about an insane need to keep sente. my only
takeaway other than reading out fights way way way in advance.

s.


On Tue, Jan 29, 2019 at 11:15 AM Gian-Carlo Pascutto  wrote:

> On 29/01/19 11:23, Petri Pitkanen wrote:
> > Just purely curiosity: How strong is Leela now? googling up gives that
> > it is better than best humasn already? Is that true?
>
> The network is over 100 Elo stronger than the second generation of ELF,
> which was about 100 Elo stronger than the first generation, which
> defeated a set of Korean top professional players 14-0.
>
> Differences in implementation speed will shift the strength difference
> around a bit, but not enough to change the conclusion that it's likely a
> lot better than the best humans now.
>
> I hear rumors it's not 100% undefeatable, and that with some trial and
> error you can occasionally still find a weaknesses to pounce on.
>
> It is used by professionals for analysis, e.g.:
> https://lifein19x19.com/viewtopic.php?f=13&t=16074
>
> --
> GCP
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaZero tensorflow implementation/tutorial

[uurtamo]

I haven't thought clearly about the 7x7 case, but on 19x19 I think it would
suffer both challenges -- you'd count dead stuff as alive quite frequently,
and because you're pruning the game ending early you might be getting wrong
who has actually won. That's why some people use less ambiguous definitions
of the end of the game.

Knowing when the game is unambiguously over and scoring it unambiguously
are good subroutines to have in your toolbelt.

s.

On Sun, Dec 9, 2018, 7:09 PM cody2007  Sorry, just to make sure I understand: your concern is the network may be
> learning from the scoring system rather than through the self-play? Or are
> you concerned the scoring is giving sub-par evaluations of games?
>
> The scoring I use is to simply count the number of stones each player has
> on the board. Then add a point for each unoccupied space that is surrounded
> completely by each player. It is simplistic and I think it does give
> sub-par evaluations of who is the winner--and definitely is a potentially
> serious deterrent to getting better performance. How much, maybe a lot.
> What do you think?
>
> ‐‐‐ Original Message ‐‐‐
> On Sunday, December 9, 2018 9:31 PM, uurtamo  wrote:
>
> Imagine that your score estimator has a better idea about the outcome of
> the game than the players themselves.
>
> Then you can build a stronger computer player with the following
> algorithm: use the score estimator to pick the next move after evaluating
> all legal moves, by evaluating their after-move scores.
>
> If you use something like Tromp-Taylor (not sure what most people use
> nowadays) then you can score it less equivocally.
>
> Perhaps I was misunderstanding, but if not then this could be a somewhat
> serious problem.
>
> s
>
>
> On Sun, Dec 9, 2018, 6:17 PM cody2007 
>> >By the way, why only 40 moves? That seems like the wrong place to
>> economize, but maybe on 7x7 it's fine?
>> I haven't implemented any resign mechanism, so felt it was a reasonable
>> balance to at least see where the players roughly stand. Although, I think
>> I errored on too few turns.
>>
>> >A "scoring estimate" by definition should be weaker than the computer
>> players it's evaluating until there are no more captures possible.
>> Not sure I understand entirely. But would agree that the scoring I use is
>> probably a limitation here.
>>
>> ‐‐‐ Original Message ‐‐‐
>> On Sunday, December 9, 2018 8:51 PM, uurtamo  wrote:
>>
>> A "scoring estimate" by definition should be weaker than the computer
>> players it's evaluating until there are no more captures possible.
>>
>> Yes?
>>
>> s.
>>
>> On Sun, Dec 9, 2018, 5:49 PM uurtamo >
>>> By the way, why only 40 moves? That seems like the wrong place to
>>> economize, but maybe on 7x7 it's fine?
>>>
>>> s.
>>>
>>> On Sun, Dec 9, 2018, 5:23 PM cody2007 via Computer-go <
>>> computer-go@computer-go.org wrote:
>>>
>>>> Thanks for your comments.
>>>>
>>>> >looks you made it work on a 7x7 19x19 would probably give better
>>>> result especially against yourself if you are a complete novice
>>>> I'd expect that'd make me win even more against the algorithm since it
>>>> would explore a far smaller amount of the search space, right?
>>>> Certainly something I'd be interested in testing though--I just would
>>>> expect it'd take many months more months of training however, but would be
>>>> interesting to see how much performance falls apart, if at all.
>>>>
>>>> >for not cheating against gnugo, use --play-out-aftermath of gnugo
>>>> parameter
>>>> Yep, I evaluate with that parameter. The problem is more that I only
>>>> play 20 turns per player per game. And the network seems to like placing
>>>> stones in terrotories "owned" by the other player. My scoring system then
>>>> no longer counts that area as owned by the player. Probably playing more
>>>> turns out and/or using a more sophisticated scoring system would fix this.
>>>>
>>>> >If I don't mistake a competitive ai would need a lot more training
>>>> such what does leela zero https://github.com/gcp/leela-zero
>>>> Yeah, I agree more training is probably the key here. I'll take a look
>>>> at leela-zero.
>>>>
>>>> ‐‐‐ Original Message ‐‐‐
>>>> On Sunday, December 9, 2018 7:41 PM, Xavier Combelle <
>>>> x

Re: [Computer-go] AlphaZero tensorflow implementation/tutorial

[uurtamo]

Imagine that your score estimator has a better idea about the outcome of
the game than the players themselves.

Then you can build a stronger computer player with the following algorithm:
use the score estimator to pick the next move after evaluating all legal
moves, by evaluating their after-move scores.

If you use something like Tromp-Taylor (not sure what most people use
nowadays) then you can score it less equivocally.

Perhaps I was misunderstanding, but if not then this could be a somewhat
serious problem.

s


On Sun, Dec 9, 2018, 6:17 PM cody2007  >By the way, why only 40 moves? That seems like the wrong place to
> economize, but maybe on 7x7 it's fine?
> I haven't implemented any resign mechanism, so felt it was a reasonable
> balance to at least see where the players roughly stand. Although, I think
> I errored on too few turns.
>
> >A "scoring estimate" by definition should be weaker than the computer
> players it's evaluating until there are no more captures possible.
> Not sure I understand entirely. But would agree that the scoring I use is
> probably a limitation here.
>
> ‐‐‐ Original Message ‐‐‐
> On Sunday, December 9, 2018 8:51 PM, uurtamo  wrote:
>
> A "scoring estimate" by definition should be weaker than the computer
> players it's evaluating until there are no more captures possible.
>
> Yes?
>
> s.
>
> On Sun, Dec 9, 2018, 5:49 PM uurtamo 
>> By the way, why only 40 moves? That seems like the wrong place to
>> economize, but maybe on 7x7 it's fine?
>>
>> s.
>>
>> On Sun, Dec 9, 2018, 5:23 PM cody2007 via Computer-go <
>> computer-go@computer-go.org wrote:
>>
>>> Thanks for your comments.
>>>
>>> >looks you made it work on a 7x7 19x19 would probably give better result
>>> especially against yourself if you are a complete novice
>>> I'd expect that'd make me win even more against the algorithm since it
>>> would explore a far smaller amount of the search space, right?
>>> Certainly something I'd be interested in testing though--I just would
>>> expect it'd take many months more months of training however, but would be
>>> interesting to see how much performance falls apart, if at all.
>>>
>>> >for not cheating against gnugo, use --play-out-aftermath of gnugo
>>> parameter
>>> Yep, I evaluate with that parameter. The problem is more that I only
>>> play 20 turns per player per game. And the network seems to like placing
>>> stones in terrotories "owned" by the other player. My scoring system then
>>> no longer counts that area as owned by the player. Probably playing more
>>> turns out and/or using a more sophisticated scoring system would fix this.
>>>
>>> >If I don't mistake a competitive ai would need a lot more training such
>>> what does leela zero https://github.com/gcp/leela-zero
>>> Yeah, I agree more training is probably the key here. I'll take a look
>>> at leela-zero.
>>>
>>> ‐‐‐ Original Message ‐‐‐
>>> On Sunday, December 9, 2018 7:41 PM, Xavier Combelle <
>>> xavier.combe...@gmail.com> wrote:
>>>
>>> looks you made it work on a 7x7 19x19 would probably give better result
>>> especially against yourself if you are a complete novice
>>>
>>> for not cheating against gnugo, use --play-out-aftermath of gnugo
>>> parameter
>>>
>>> If I don't mistake a competitive ai would need a lot more training such
>>> what does leela zero https://github.com/gcp/leela-zero
>>> Le 10/12/2018 à 01:25, cody2007 via Computer-go a écrit :
>>>
>>> Hi all,
>>>
>>> I've posted an implementation of the AlphaZero algorithm and brief
>>> tutorial. The code runs on a single GPU. While performance is not that
>>> great, I suspect its mostly been limited by hardware limitations (my
>>> training and evaluation has been on a single Titan X). The network can beat
>>> GNU go about 50% of the time, although it "abuses" the scoring a little
>>> bit--which I talk a little more about in the article:
>>>
>>>
>>> https://medium.com/@cody2007.2/alphazero-implementation-and-tutorial-f4324d65fdfc
>>>
>>> -Cody
>>>
>>> ___
>>> Computer-go mailing 
>>> listComputer-go@computer-go.orghttp://computer-go.org/mailman/listinfo/computer-go
>>>
>>>
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaZero tensorflow implementation/tutorial

[uurtamo]

A "scoring estimate" by definition should be weaker than the computer
players it's evaluating until there are no more captures possible.

Yes?

s.

On Sun, Dec 9, 2018, 5:49 PM uurtamo  By the way, why only 40 moves? That seems like the wrong place to
> economize, but maybe on 7x7 it's fine?
>
> s.
>
> On Sun, Dec 9, 2018, 5:23 PM cody2007 via Computer-go <
> computer-go@computer-go.org wrote:
>
>> Thanks for your comments.
>>
>> >looks you made it work on a 7x7 19x19 would probably give better result
>> especially against yourself if you are a complete novice
>> I'd expect that'd make me win even more against the algorithm since it
>> would explore a far smaller amount of the search space, right?
>> Certainly something I'd be interested in testing though--I just would
>> expect it'd take many months more months of training however, but would be
>> interesting to see how much performance falls apart, if at all.
>>
>> >for not cheating against gnugo, use --play-out-aftermath of gnugo
>> parameter
>> Yep, I evaluate with that parameter. The problem is more that I only play
>> 20 turns per player per game. And the network seems to like placing stones
>> in terrotories "owned" by the other player. My scoring system then no
>> longer counts that area as owned by the player. Probably playing more turns
>> out and/or using a more sophisticated scoring system would fix this.
>>
>> >If I don't mistake a competitive ai would need a lot more training such
>> what does leela zero https://github.com/gcp/leela-zero
>> Yeah, I agree more training is probably the key here. I'll take a look at
>> leela-zero.
>>
>> ‐‐‐ Original Message ‐‐‐
>> On Sunday, December 9, 2018 7:41 PM, Xavier Combelle <
>> xavier.combe...@gmail.com> wrote:
>>
>> looks you made it work on a 7x7 19x19 would probably give better result
>> especially against yourself if you are a complete novice
>>
>> for not cheating against gnugo, use --play-out-aftermath of gnugo
>> parameter
>>
>> If I don't mistake a competitive ai would need a lot more training such
>> what does leela zero https://github.com/gcp/leela-zero
>> Le 10/12/2018 à 01:25, cody2007 via Computer-go a écrit :
>>
>> Hi all,
>>
>> I've posted an implementation of the AlphaZero algorithm and brief
>> tutorial. The code runs on a single GPU. While performance is not that
>> great, I suspect its mostly been limited by hardware limitations (my
>> training and evaluation has been on a single Titan X). The network can beat
>> GNU go about 50% of the time, although it "abuses" the scoring a little
>> bit--which I talk a little more about in the article:
>>
>>
>> https://medium.com/@cody2007.2/alphazero-implementation-and-tutorial-f4324d65fdfc
>>
>> -Cody
>>
>> ___
>> Computer-go mailing 
>> listComputer-go@computer-go.orghttp://computer-go.org/mailman/listinfo/computer-go
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaZero tensorflow implementation/tutorial

[uurtamo]

By the way, why only 40 moves? That seems like the wrong place to
economize, but maybe on 7x7 it's fine?

s.

On Sun, Dec 9, 2018, 5:23 PM cody2007 via Computer-go <
computer-go@computer-go.org wrote:

> Thanks for your comments.
>
> >looks you made it work on a 7x7 19x19 would probably give better result
> especially against yourself if you are a complete novice
> I'd expect that'd make me win even more against the algorithm since it
> would explore a far smaller amount of the search space, right?
> Certainly something I'd be interested in testing though--I just would
> expect it'd take many months more months of training however, but would be
> interesting to see how much performance falls apart, if at all.
>
> >for not cheating against gnugo, use --play-out-aftermath of gnugo
> parameter
> Yep, I evaluate with that parameter. The problem is more that I only play
> 20 turns per player per game. And the network seems to like placing stones
> in terrotories "owned" by the other player. My scoring system then no
> longer counts that area as owned by the player. Probably playing more turns
> out and/or using a more sophisticated scoring system would fix this.
>
> >If I don't mistake a competitive ai would need a lot more training such
> what does leela zero https://github.com/gcp/leela-zero
> Yeah, I agree more training is probably the key here. I'll take a look at
> leela-zero.
>
> ‐‐‐ Original Message ‐‐‐
> On Sunday, December 9, 2018 7:41 PM, Xavier Combelle <
> xavier.combe...@gmail.com> wrote:
>
> looks you made it work on a 7x7 19x19 would probably give better result
> especially against yourself if you are a complete novice
>
> for not cheating against gnugo, use --play-out-aftermath of gnugo parameter
>
> If I don't mistake a competitive ai would need a lot more training such
> what does leela zero https://github.com/gcp/leela-zero
> Le 10/12/2018 à 01:25, cody2007 via Computer-go a écrit :
>
> Hi all,
>
> I've posted an implementation of the AlphaZero algorithm and brief
> tutorial. The code runs on a single GPU. While performance is not that
> great, I suspect its mostly been limited by hardware limitations (my
> training and evaluation has been on a single Titan X). The network can beat
> GNU go about 50% of the time, although it "abuses" the scoring a little
> bit--which I talk a little more about in the article:
>
>
> https://medium.com/@cody2007.2/alphazero-implementation-and-tutorial-f4324d65fdfc
>
> -Cody
>
> ___
> Computer-go mailing 
> listComputer-go@computer-go.orghttp://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] New paper by DeepMind

[uurtamo]

So published prior art isn't a defense? It's pretty widely publicized what
they did and how.

The problem I have with most tech patents is when they're overly broad.

s.

On Sun, Dec 9, 2018, 9:11 AM David Doshay via Computer-go <
computer-go@computer-go.org wrote:

> Another very important aspect of this discussion is that the US patent
> office changed to a ‘first to file’ method of prioritizing patent rights.
> This encouraged several patent trolls to try to undercut the true
> inventors. So, it is now more important to file for defensive purposes just
> to assure that deep pockets like Alpha do not have to pay royalties to
> others for their own inventions.
>
> Many years ago when I worked at NASA we were researching doing a patent
> filing for an image processing technique so that we could release it for
> public domain use. We found that someone successfully got a patent for
> using a bitmap to represent a black-and-white image! It may indeed have
> been possible and successful to argue in court that this is obvious to
> anyone in the industry and thus should not be granted a patent, but it
> would be costly and a bother to have to do so. Likewise for a deep pocket
> like Alpha who would be an obvious target for patent trolling if they did
> not get this technique labeled as public knowledge quickly enough.
>
> Cheers,
> David
>
> On Dec 9, 2018, at 8:30 AM, Jim O'Flaherty 
> wrote:
>
> Tysvm for your excellent explanation.
>
> And now you can see why I mentioned Google's being a member of OIN as a
> critical distinction. It strongly increases the weight of 2. And implicitly
> reduces the motivation for 1.
>
>
> On Sat, Dec 8, 2018, 8:51 PM 甲斐徳本 
>> Those are the points not well understood commonly.
>>
>> A patent application does two things.  1. Apply for an eventual granting
>> of the patent, 2. Makes what's described in it a public knowledge as of the
>> date of the filing.
>> Patent may be functionally meaningless.  There may be no one to sue.  And
>> these are huge issues for the point No.1.  However, a strategic patent
>> applicants file patent applications for the point No.2 to deny any
>> possibility of somebody else obtaining a patent.  (A public knowledge
>> cannot be patented.)
>>
>> Many companies are trying to figure out how to patent DCNN based AI, and
>> Google may be saying "Nope, as long as it is like the DeepMind method, you
>> can't patent it."   Google is likely NOT saying "We are hoping to obtain
>> the patent, and intend to enforce it."
>>
>> Despite many differences in patent law from a country to another, two
>> basic purposes of patent are universal: 1. To protect the inventor, and 2.
>> To promote the use of inventions by making the details a public knowledge.
>>
>>
>>
>>
>> On Sat, Dec 8, 2018 at 12:47 AM uurtamo  wrote:
>>
>>> What I'm saying is that the patent is functionally meaningless. Who is
>>> there to sue?
>>>
>>> Moreover, there is no enforceable patent on the broad class of
>>> algorithms that could reproduce these results. No?
>>>
>>> s.
>>>
>>> On Fri, Dec 7, 2018, 4:16 AM Jim O'Flaherty >> wrote:
>>>
>>>> Tysvm for the clarification, Tokumoto.
>>>>
>>>> On Thu, Dec 6, 2018, 8:02 PM 甲斐徳本 >>>
>>>>> What's insane about it?
>>>>> To me, what Jim O'Flaherty stated is common sense in the field of
>>>>> patents, and any patent attorney would attest to that.  If I may add, 
>>>>> Jim's
>>>>> last sentence should read "Google's patent application" instead of
>>>>> "Google's patent".  The difference is huge, and this may be in the heart 
>>>>> of
>>>>> the issue, which is not well understood by the general public.
>>>>>
>>>>> In other words, thousands of patent applications are filed in the
>>>>> world without any hope of the patent eventually being granted, to 
>>>>> establish
>>>>> "prior art" thereby protecting what's described in it from being patented
>>>>> by somebody else.
>>>>>
>>>>> Or, am I responding to a troll?
>>>>>
>>>>> Tokumoto
>>>>>
>>>>>
>>>>> On Fri, Dec 7, 2018 at 10:01 AM uurtamo  wrote:
>>>>>
>>>>>> You're insane.
>>>>>>
>>>>>> On Thu, Dec 6, 2018, 4:13 PM 

Re: [Computer-go] New paper by DeepMind

[uurtamo]

Thank you for this clarification,

s.

On Sat, Dec 8, 2018, 6:51 PM 甲斐徳本  Those are the points not well understood commonly.
>
> A patent application does two things.  1. Apply for an eventual granting
> of the patent, 2. Makes what's described in it a public knowledge as of the
> date of the filing.
> Patent may be functionally meaningless.  There may be no one to sue.  And
> these are huge issues for the point No.1.  However, a strategic patent
> applicants file patent applications for the point No.2 to deny any
> possibility of somebody else obtaining a patent.  (A public knowledge
> cannot be patented.)
>
> Many companies are trying to figure out how to patent DCNN based AI, and
> Google may be saying "Nope, as long as it is like the DeepMind method, you
> can't patent it."   Google is likely NOT saying "We are hoping to obtain
> the patent, and intend to enforce it."
>
> Despite many differences in patent law from a country to another, two
> basic purposes of patent are universal: 1. To protect the inventor, and 2.
> To promote the use of inventions by making the details a public knowledge.
>
>
>
>
> On Sat, Dec 8, 2018 at 12:47 AM uurtamo  wrote:
>
>> What I'm saying is that the patent is functionally meaningless. Who is
>> there to sue?
>>
>> Moreover, there is no enforceable patent on the broad class of algorithms
>> that could reproduce these results. No?
>>
>> s.
>>
>> On Fri, Dec 7, 2018, 4:16 AM Jim O'Flaherty > wrote:
>>
>>> Tysvm for the clarification, Tokumoto.
>>>
>>> On Thu, Dec 6, 2018, 8:02 PM 甲斐徳本 >>
>>>> What's insane about it?
>>>> To me, what Jim O'Flaherty stated is common sense in the field of
>>>> patents, and any patent attorney would attest to that.  If I may add, Jim's
>>>> last sentence should read "Google's patent application" instead of
>>>> "Google's patent".  The difference is huge, and this may be in the heart of
>>>> the issue, which is not well understood by the general public.
>>>>
>>>> In other words, thousands of patent applications are filed in the world
>>>> without any hope of the patent eventually being granted, to establish
>>>> "prior art" thereby protecting what's described in it from being patented
>>>> by somebody else.
>>>>
>>>> Or, am I responding to a troll?
>>>>
>>>> Tokumoto
>>>>
>>>>
>>>> On Fri, Dec 7, 2018 at 10:01 AM uurtamo  wrote:
>>>>
>>>>> You're insane.
>>>>>
>>>>> On Thu, Dec 6, 2018, 4:13 PM Jim O'Flaherty <
>>>>> jim.oflaherty...@gmail.com wrote:
>>>>>
>>>>>> Remember, patents are a STRATEGIC mechanism as well as a legal
>>>>>> mechanism. As soon as a patent is publically filed (for example, as
>>>>>> utility, and following provisional), the text and claims in the patent
>>>>>> immediately become prior art globally as of the original filing date
>>>>>> REGARDLESS of whether the patent is eventually approved or rejected. 
>>>>>> IOW, a
>>>>>> patent filing is a mechanism to ensure no one else can make a similar 
>>>>>> claim
>>>>>> without risking this filing being used as a possible prior art 
>>>>>> refutation.
>>>>>>
>>>>>> I know this only because it is a strategy option my company is using
>>>>>> in an entirely different unrelated domain. The patent filing is defensive
>>>>>> such that someone else cannot make a claim and take our inventions away
>>>>>> from us just because the coincidentally hit near our inventions.
>>>>>>
>>>>>> So considering Google's past and their participation in the OIN, it
>>>>>> is very likely Google's patent is ensuring the ground all around this 
>>>>>> area
>>>>>> is sufficiently salted to stop anyone from attempting to exploit nearby
>>>>>> patent claims.
>>>>>>
>>>>>>
>>>>>> Respectfully,
>>>>>>
>>>>>> Jim O'Flaherty
>>>>>>
>>>>>>
>>>>>> On Thu, Dec 6, 2018 at 5:44 PM Erik van der Werf <
>>>>>> erikvanderw...@gmail.com>

Re: [Computer-go] New paper by DeepMind

[uurtamo]

What I'm saying is that the patent is functionally meaningless. Who is
there to sue?

Moreover, there is no enforceable patent on the broad class of algorithms
that could reproduce these results. No?

s.

On Fri, Dec 7, 2018, 4:16 AM Jim O'Flaherty  Tysvm for the clarification, Tokumoto.
>
> On Thu, Dec 6, 2018, 8:02 PM 甲斐徳本 
>> What's insane about it?
>> To me, what Jim O'Flaherty stated is common sense in the field of
>> patents, and any patent attorney would attest to that.  If I may add, Jim's
>> last sentence should read "Google's patent application" instead of
>> "Google's patent".  The difference is huge, and this may be in the heart of
>> the issue, which is not well understood by the general public.
>>
>> In other words, thousands of patent applications are filed in the world
>> without any hope of the patent eventually being granted, to establish
>> "prior art" thereby protecting what's described in it from being patented
>> by somebody else.
>>
>> Or, am I responding to a troll?
>>
>> Tokumoto
>>
>>
>> On Fri, Dec 7, 2018 at 10:01 AM uurtamo  wrote:
>>
>>> You're insane.
>>>
>>> On Thu, Dec 6, 2018, 4:13 PM Jim O'Flaherty >> wrote:
>>>
>>>> Remember, patents are a STRATEGIC mechanism as well as a legal
>>>> mechanism. As soon as a patent is publically filed (for example, as
>>>> utility, and following provisional), the text and claims in the patent
>>>> immediately become prior art globally as of the original filing date
>>>> REGARDLESS of whether the patent is eventually approved or rejected. IOW, a
>>>> patent filing is a mechanism to ensure no one else can make a similar claim
>>>> without risking this filing being used as a possible prior art refutation.
>>>>
>>>> I know this only because it is a strategy option my company is using in
>>>> an entirely different unrelated domain. The patent filing is defensive such
>>>> that someone else cannot make a claim and take our inventions away from us
>>>> just because the coincidentally hit near our inventions.
>>>>
>>>> So considering Google's past and their participation in the OIN, it is
>>>> very likely Google's patent is ensuring the ground all around this area is
>>>> sufficiently salted to stop anyone from attempting to exploit nearby patent
>>>> claims.
>>>>
>>>>
>>>> Respectfully,
>>>>
>>>> Jim O'Flaherty
>>>>
>>>>
>>>> On Thu, Dec 6, 2018 at 5:44 PM Erik van der Werf <
>>>> erikvanderw...@gmail.com> wrote:
>>>>
>>>>> On Thu, Dec 6, 2018 at 11:28 PM Rémi Coulom 
>>>>> wrote:
>>>>>
>>>>>> Also, the AlphaZero algorithm is patented:
>>>>>> https://patentscope2.wipo.int/search/en/detail.jsf?docId=WO2018215665
>>>>>>
>>>>>
>>>>> So far it just looks like an application (and I don't think it will be
>>>>> be difficult to oppose, if you care about this)
>>>>>
>>>>> Erik
>>>>>
>>>>> ___
>>>>> Computer-go mailing list
>>>>> Computer-go@computer-go.org
>>>>> http://computer-go.org/mailman/listinfo/computer-go
>>>>
>>>> ___
>>>> Computer-go mailing list
>>>> Computer-go@computer-go.org
>>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] New paper by DeepMind

[uurtamo]

You're insane.

On Thu, Dec 6, 2018, 4:13 PM Jim O'Flaherty  Remember, patents are a STRATEGIC mechanism as well as a legal mechanism.
> As soon as a patent is publically filed (for example, as utility, and
> following provisional), the text and claims in the patent immediately
> become prior art globally as of the original filing date REGARDLESS of
> whether the patent is eventually approved or rejected. IOW, a patent filing
> is a mechanism to ensure no one else can make a similar claim without
> risking this filing being used as a possible prior art refutation.
>
> I know this only because it is a strategy option my company is using in an
> entirely different unrelated domain. The patent filing is defensive such
> that someone else cannot make a claim and take our inventions away from us
> just because the coincidentally hit near our inventions.
>
> So considering Google's past and their participation in the OIN, it is
> very likely Google's patent is ensuring the ground all around this area is
> sufficiently salted to stop anyone from attempting to exploit nearby patent
> claims.
>
>
> Respectfully,
>
> Jim O'Flaherty
>
>
> On Thu, Dec 6, 2018 at 5:44 PM Erik van der Werf 
> wrote:
>
>> On Thu, Dec 6, 2018 at 11:28 PM Rémi Coulom  wrote:
>>
>>> Also, the AlphaZero algorithm is patented:
>>> https://patentscope2.wipo.int/search/en/detail.jsf?docId=WO2018215665
>>>
>>
>> So far it just looks like an application (and I don't think it will be be
>> difficult to oppose, if you care about this)
>>
>> Erik
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

Did not intend to go to the whole group

On Thu, Jun 21, 2018, 5:01 PM uurtamo  wrote:

> So deep down I think that first capture isn't that hard.
>
> I also think that what makes real go that hard is ko, but you've shown
> that it's equivalent to ladder, which frankly baffles me. I'd love to
> understand that.
>
> You've done great combinatorics work and great small scale work.
>
> What's your thinking about interesting problems forward?
>
> s.
>
> On Thu, Jun 21, 2018, 4:52 PM John Tromp  wrote:
>
>> >>> Direct link to image: http://tromp.github.io/img/WO5lives.png
>>
>> Might be useful for go event organizers in need of arrow signs...
>>
>> regards,
>> -John
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

So deep down I think that first capture isn't that hard.

I also think that what makes real go that hard is ko, but you've shown that
it's equivalent to ladder, which frankly baffles me. I'd love to understand
that.

You've done great combinatorics work and great small scale work.

What's your thinking about interesting problems forward?

s.

On Thu, Jun 21, 2018, 4:52 PM John Tromp  wrote:

> >>> Direct link to image: http://tromp.github.io/img/WO5lives.png
>
> Might be useful for go event organizers in need of arrow signs...
>
> regards,
> -John
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

Re: trolling

On Thu, Jun 21, 2018, 4:37 PM Mario Xerxes Castelán Castro <
marioxcc...@yandex.com> wrote:

> “He” is the genetic singular pronoun in English. If anybody feels
> excluded, is because he wants to feel excluded or is intentionally
> playing the ignorant card. What happen is that the social justice
> warrior-dominated United States is the source of many attempts to
> redefine reality when it goes against the ultraliberal agenda.
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

Without discouraging speech of any kind, I'd like to suggest that the prior
statement was of the form "axe to grind" or "mild trolling".

s.

On Thu, Jun 21, 2018, 1:45 PM Dan Schmidt  wrote:

>
> On Thu, Jun 21, 2018 at 1:41 PM, Mario Xerxes Castelán Castro <
> marioxcc...@yandex.com> wrote:
>
> Why the misandry? In English, “he” serves for both neutral and male
>> gender, but “she” always excludes men.
>>
>
> Standards change. I could give examples of constructions that used to be
> considered polite but are no longer, but by definition they wouldn't be
> polite...
>
> I'm not going to get into an argument about usage (and I hope no one else
> here is either, as that is not the subject of this mailing list), so this
> will be my only reply on the subject, but here is one starting point if you
> are interested in the topic, as you seem to be:
> https://www.apaonline.org/page/nonsexist (the American Philosophical
> Association's "Guidelines for Non-Sexist Use Of Language"). Searching for
> "gender-fair language" on the internet will turn up plenty of other
> resources.
>
> Dan
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

Understood, and thanks. I didn't mean to throw the conversation sideways
too much.

Steve


On Tue, Jun 19, 2018 at 7:22 AM Marcel Crasmaru  wrote:

> > _first capture_, no?
>
> I think there is some misunderstanding here as in this thread several
> problems are discussed in parallel.
>
> If by _first capture_ you mean to find an answer to the question "what
> is the computational difficulty of Capture GO?" then as far as I know
> no one proved anything yet. Capture GO might be in P but to prove this
> doesn't look like an easy task. I personally think it is either
>
> (1) in P but very hard to prove it, or
> (2) at least NP hard because, empirically, you may still create few
> convoluted ladders that don't capture stones and interact to each
> other in unexpected ways etc. Using loose ladders might be a another
> way to try building NP hard instances. However, without a proof this
> assumption is still as valid as (1).
>
> I am curious what's John Tromp opinion on the above.
>
> (Please note that the problem I've created has nothing to do with Capture
> GO.)
>
> Thanks,
> Marcel
>
> On 19 June 2018 at 13:10, uurtamo  wrote:
> > _first capture_, no?
> >
> > s.
> >
> > On Mon, Jun 18, 2018, 6:59 PM Marcel Crasmaru 
> wrote:
> >>
> >> I've eventually managed to create a problem that should show a full
> >> reduction from a Robson problem to Go - I hope is correct.
> >>
> >> The Problem:
> >>
> https://drive.google.com/file/d/1tmClDIs-baXUqRC7fQ2iKzMRXoQuGmz2/view?usp=sharing
> >> Black just captured in the marked ko. How should White play to save
> >> the lower group?
> >>
> >>
> >> > no ko fights and no counting (i.e. first capture) could put this in P.
> >>
> >> Not true - please read Tromp et al paper: Ladders are PSPACE hard
> >> without ko - that is, you can reduce any PSPACE problem in reasonable
> >> time to a Go problem without kos.
> >>
> >> --Marcel
> >>
> >> On 18 June 2018 at 22:27, uurtamo  wrote:
> >> > My understanding: ko fights will take this to (at least, I haven't
> seen
> >> > the
> >> > EXP argument) PSPACE.
> >> >
> >> > no ko fights and no counting (i.e. first capture) could put this in P.
> >> >
> >> > s.
> >> >
> >> >
> >> > On Mon, Jun 18, 2018 at 3:21 PM John Tromp 
> wrote:
> >> >>
> >> >> On Mon, Jun 18, 2018 at 10:24 PM, Álvaro Begué <
> alvaro.be...@gmail.com>
> >> >> wrote:
> >> >> > I don't think ko fights have anything to do with this. John Tromp
> >> >> > told
> >> >> > me that ladders are PSPACE complete:
> https://tromp.github.io/lad.ps
> >> >>
> >> >> Ko fights are needed to take Go problems beyond PSPACE.
> >> >> For Japanese rules they suffice to go beyond (assuming EXPTIME !=
> >> >> PSPACE),
> >> >> but for Chinese rules it's an open problem.
> >> >>
> >> >> regards,
> >> >> -John
> >> >> ___
> >> >> Computer-go mailing list
> >> >> Computer-go@computer-go.org
> >> >> http://computer-go.org/mailman/listinfo/computer-go
> >> >
> >> >
> >> > ___
> >> > Computer-go mailing list
> >> > Computer-go@computer-go.org
> >> > http://computer-go.org/mailman/listinfo/computer-go
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> >
> >
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

_first capture_, no?

s.

On Mon, Jun 18, 2018, 6:59 PM Marcel Crasmaru  wrote:

> I've eventually managed to create a problem that should show a full
> reduction from a Robson problem to Go - I hope is correct.
>
> The Problem:
> https://drive.google.com/file/d/1tmClDIs-baXUqRC7fQ2iKzMRXoQuGmz2/view?usp=sharing
> Black just captured in the marked ko. How should White play to save
> the lower group?
>
>
> > no ko fights and no counting (i.e. first capture) could put this in P.
>
> Not true - please read Tromp et al paper: Ladders are PSPACE hard
> without ko - that is, you can reduce any PSPACE problem in reasonable
> time to a Go problem without kos.
>
> --Marcel
>
> On 18 June 2018 at 22:27, uurtamo  wrote:
> > My understanding: ko fights will take this to (at least, I haven't seen
> the
> > EXP argument) PSPACE.
> >
> > no ko fights and no counting (i.e. first capture) could put this in P.
> >
> > s.
> >
> >
> > On Mon, Jun 18, 2018 at 3:21 PM John Tromp  wrote:
> >>
> >> On Mon, Jun 18, 2018 at 10:24 PM, Álvaro Begué 
> >> wrote:
> >> > I don't think ko fights have anything to do with this. John Tromp told
> >> > me that ladders are PSPACE complete: https://tromp.github.io/lad.ps
> >>
> >> Ko fights are needed to take Go problems beyond PSPACE.
> >> For Japanese rules they suffice to go beyond (assuming EXPTIME !=
> PSPACE),
> >> but for Chinese rules it's an open problem.
> >>
> >> regards,
> >> -John
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> >
> >
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

My understanding: ko fights will take this to (at least, I haven't seen the
EXP argument) PSPACE.

no ko fights and no counting (i.e. first capture) could put this in P.

s.


On Mon, Jun 18, 2018 at 3:21 PM John Tromp  wrote:

> On Mon, Jun 18, 2018 at 10:24 PM, Álvaro Begué 
> wrote:
> > I don't think ko fights have anything to do with this. John Tromp told
> > me that ladders are PSPACE complete: https://tromp.github.io/lad.ps
>
> Ko fights are needed to take Go problems beyond PSPACE.
> For Japanese rules they suffice to go beyond (assuming EXPTIME != PSPACE),
> but for Chinese rules it's an open problem.
>
> regards,
> -John
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

Ko fights are the thing. Ladders are hard, but without ko fights I'm pretty
sure it's not even PSPACE-complete.

Steve


On Mon, Jun 18, 2018 at 1:52 PM Álvaro Begué  wrote:

> I don't think ko fights have anything to do with this. John Tromp told
> me that ladders are PSPACE complete: https://tromp.github.io/lad.ps
>
> Álvaro.
>
>
>
> On Mon, Jun 18, 2018 at 2:58 PM, uurtamo  wrote:
> > FWIW, first-capture go (i.e. winner is first one to make a capture)
> should
> > not be PSPACE-complete.
> >
> > the thing in go that makes it hard is ko fights, which don't exist in
> > capture go.
> >
> > s.
> >
> >
> > On Mon, Jun 18, 2018 at 11:55 AM Marcel Crasmaru 
> > wrote:
> >>
> >> Errata: > reduction from GO to an EXP hard problem
> >>
> >> should be the other way around :)
> >>
> >> --Marcel
> >>
> >> On 18 June 2018 at 19:36, Marcel Crasmaru  wrote:
> >> >>   J. M. Robson (1983) “The Complexity of Go”. Proceedings of the IFIP
> >> >> Congress 1983 p. 413-417.
> >> >
> >> > If you are interested in how to prove that GO with kos and Japanese
> >> > rules is EXP complete you can get the gist of it from a very early
> >> > draft of my master thesis
> >> > - I used Robson's idea of reduction from GO to an EXP hard problem
> >> > using ladders instead of pipes (he used groups
> >> > connected through long string of pieces, aka, "pipes")
> >> >
> >> > If you have related questions I am happy to answer them although John
> >> > Tromp might have even better insights - ask him too.
> >> >
> >> > Best,
> >> > Marcel
> >> >
> >> > On 18 June 2018 at 17:54, Mario Xerxes Castelán Castro
> >> >  wrote:
> >> >> Hello. I am asking for help finding the following paper:
> >> >>
> >> >>   J. M. Robson (1983) “The Complexity of Go”. Proceedings of the IFIP
> >> >> Congress 1983 p. 413-417.
> >> >>
> >> >> I could not find it online. There is no DOI anywhere to be found (I
> >> >> searched Crossref and here:
> >> >> https://dblp.uni-trier.de/db/conf/ifip/ifip83.html#Robson83 ) and
> the
> >> >> conference proceedings are not in Library Genesis either.
> >> >>
> >> >> Thanks in advance.
> >> >>
> >> >>
> >> >> ___
> >> >> Computer-go mailing list
> >> >> Computer-go@computer-go.org
> >> >> http://computer-go.org/mailman/listinfo/computer-go
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> >
> >
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Paper “Complexity of Go” by Robson

[uurtamo]

FWIW, first-capture go (i.e. winner is first one to make a capture) should
not be PSPACE-complete.

the thing in go that makes it hard is ko fights, which don't exist in
capture go.

s.


On Mon, Jun 18, 2018 at 11:55 AM Marcel Crasmaru 
wrote:

> Errata: > reduction from GO to an EXP hard problem
>
> should be the other way around :)
>
> --Marcel
>
> On 18 June 2018 at 19:36, Marcel Crasmaru  wrote:
> >>   J. M. Robson (1983) “The Complexity of Go”. Proceedings of the IFIP
> Congress 1983 p. 413-417.
> >
> > If you are interested in how to prove that GO with kos and Japanese
> > rules is EXP complete you can get the gist of it from a very early
> > draft of my master thesis
> > - I used Robson's idea of reduction from GO to an EXP hard problem
> > using ladders instead of pipes (he used groups
> > connected through long string of pieces, aka, "pipes")
> >
> > If you have related questions I am happy to answer them although John
> > Tromp might have even better insights - ask him too.
> >
> > Best,
> > Marcel
> >
> > On 18 June 2018 at 17:54, Mario Xerxes Castelán Castro
> >  wrote:
> >> Hello. I am asking for help finding the following paper:
> >>
> >>   J. M. Robson (1983) “The Complexity of Go”. Proceedings of the IFIP
> >> Congress 1983 p. 413-417.
> >>
> >> I could not find it online. There is no DOI anywhere to be found (I
> >> searched Crossref and here:
> >> https://dblp.uni-trier.de/db/conf/ifip/ifip83.html#Robson83 ) and the
> >> conference proceedings are not in Library Genesis either.
> >>
> >> Thanks in advance.
> >>
> >>
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] CGOS Real-time game viewer

[uurtamo]

by the way, i'm watching LZ get killed in the early game by something
called hd? who is that?

at the end it's all fairly sloppy (this only based upon 3-4 games i've
watched), but hd seems to have a strong opening comparatively.

s.


On Wed, May 30, 2018 at 6:04 AM, David Wu  wrote:

> Awesome!
>
> I was going to complain that the color scheme had not enough contrast
> between the white stones and the background (at least on my desktop, it was
> hard to see the white stones clearly), and then I discovered the settings
> menu already lets you change the background color. Cool! :)
>
>
> On Wed, May 30, 2018 at 8:38 AM, uurtamo  wrote:
>
>> This is really well done.
>>
>> Thanks,
>>
>> Steve
>>
>>
>> On Tue, May 29, 2018 at 4:10 PM, Hiroshi Yamashita 
>> wrote:
>>
>>> Hi,
>>>
>>> CGOS 19x19 Real-time game viewer is available.
>>> https://deepleela.com/cgos
>>>
>>> Thank you for author of DeepLeela site.
>>> DeepLeela logins as a Viewer, and dispatches to thier clients.
>>> There were Viewing Client on Linux and Windows, but no on the browser.
>>>
>>> Source code is also available.
>>> https://github.com/deepleela
>>>
>>> Thanks,
>>> Hiroshi Yamashita
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] CGOS Real-time game viewer

[uurtamo]

This is really well done.

Thanks,

Steve


On Tue, May 29, 2018 at 4:10 PM, Hiroshi Yamashita  wrote:

> Hi,
>
> CGOS 19x19 Real-time game viewer is available.
> https://deepleela.com/cgos
>
> Thank you for author of DeepLeela site.
> DeepLeela logins as a Viewer, and dispatches to thier clients.
> There were Viewing Client on Linux and Windows, but no on the browser.
>
> Source code is also available.
> https://github.com/deepleela
>
> Thanks,
> Hiroshi Yamashita
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] 9x9 is last frontier?

[uurtamo .]

Summarizing the objections to my (non-evidence-based, but hand-wavy
observationally-based) assertion that 9x9 is going down anytime someone
really wants it to go down, I get the following:

* value networks can't hack it (okay, maybe? does this make it less likely?
-- we shouldn't expect to cut-and-paste.)
* double ko is some kind of super special problem on 9x9
* the margins are way narrower

the second issue (ko in general, but multi-ko) is exactly why go is
pspace-complete (I made a rough argument for the fact that under slight
relaxation it isn't, but didn't flesh it out or publish it).

I am not feeling strong arguments about the overall picture (i.e. it's
super much harder than 19x19 to beat humans at) other than that the margins
are narrower.

Does anyone else have a different synopsis?

Thanks,

Steve




On Tue, Mar 6, 2018 at 12:17 PM, Brian Sheppard via Computer-go <
computer-go@computer-go.org> wrote:

> Well, AlphaZero did fine at chess tactics, and the papers are clear on the
> details. There must be an error in your deductions somewhere.
>
>
>
> *From:* Computer-go [mailto:computer-go-boun...@computer-go.org] *On
> Behalf Of *Dan
> *Sent:* Tuesday, March 6, 2018 1:46 PM
>
> *To:* computer-go@computer-go.org
> *Subject:* Re: [Computer-go] 9x9 is last frontier?
>
>
>
> I am pretty sure it is an MCTS problem and I suspect not something that
> could be easily solved with a policy network (could be wrong hree). My
> opinon is that DCNN is not
>
> a miracle worker (as somebody already mentioned here) and it is going to
> fail  resolving tactics.  I would be more than happy with it if it has same
> power as a qsearch to be honest.
>
>
>
> Search traps are the major problem with games like Chess, and what makes
> transitioning the success of DCNN from Go to Chess non trivial.
>
> The following paper discusses shallow traps that are prevalent in chess. (
> https://www.aaai.org/ocs/index.php/ICAPS/ICAPS10/paper/download/1458/1571
> )
>
> They mention traps make MCTS very inefficient.  Even if the MCTS is given
> 50x more time is needed by an exhaustive minimax tree, it could fail to
> find a level-5 or level-7 trap.
>
> It will spend, f.i, 95% of its time searching an asymetric tree of depth >
> 7 when a shallow trap of depth-7 exists, thus, missing to find the level-7
> trap.
>
> This is very hard to solve even if you have unlimited power.
>
>
>
> The plain MCTS as used by AlphaZero is the most ill-suited MCTS version in
> my opinion and i have hard a hard time seeing how it can be competitive
> with Stockfish tactically.
>
>
>
> My MCTS chess engine with  AlphaZero like MCTS was averaging was missing a
> lot of tactics. I don't use policy or eval networks but qsearch() for eval,
> and the policy is basically
>
> choosing which ever moves leads to a higher eval.
>
>
>
> a) My first improvement to the MCTS is to use minimax backups instead of
> averaging. This was an improvmenet but not something that would solve the
> traps
>
>
>
> b) My second improvment is to use alphabeta rollouts. This is a rollouts
> version that can do nullmove and LMR etc... This is a huge improvment and
> none of the MCTS
>
> versons can match it. More on alpha-beta rollouts here (
> https://www.microsoft.com/en-us/research/wp-content/
> uploads/2014/11/huang_rollout.pdf )
>
>
>
> So AlphaZero used none of the above improvements and yet it seems to be
> tactically strong. Leela-Zero suffered from tactical falls left and right
> too as I expected.
>
>
>
> So the only explanation left is the policy network able to avoid traps
> which I find hard to believe it can identify more than a qsearch level
> tactics.
>
>
>
> All I am saying is that my experience (as well as many others) with MCTS
> for tactical dominated games is bad, and there must be some breakthrough in
> that regard in AlphaZero
>
> for it to be able to compete with Stockfish on a tactical level.
>
>
>
> I am curious how Remi's attempt at Shogi using AlphaZero's method will
> turnout.
>
>
>
> regards,
>
> Daniel
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> On Tue, Mar 6, 2018 at 9:41 AM, Brian Sheppard via Computer-go <
> computer-go@computer-go.org> wrote:
>
> Training on Stockfish games is guaranteed to produce a blunder-fest,
> because there are no blunders in the training set and therefore the policy
> network never learns how to refute blunders.
>
>
>
> This is not a flaw in MCTS, but rather in the policy network. MCTS will
> eventually search every move infinitely often, producing asymptotically
> optimal play. But if the policy network does not provide the guidance
> necessary to rapidly refute the blunders that occur in the search, then
> convergence of MCTS to optimal play will be very slow.
>
>
>
> It is necessary for the network to train on self-play games using MCTS.
> For instance, the AGZ approach samples next states during training games by
> sampling from the distribution of visits in the search. Specifically: not
> by choosing the most-visited play!
>
>
>
> 

Re: [Computer-go] 9x9 is last frontier?

[uurtamo .]

Thank you for being so kind in your response. I truly appreciate it.

s.

On Feb 28, 2018 6:32 PM, "Hideki Kato"  wrote:

> uurtamo .:  mail.gmail.com>:
> >I didn't mean to suggest that I can or will solve this problem tomorrow.
> >
> >What I meant to say is that it is clearly obvious that 9x9 is not immune
> to
> >being destroyed -- it's not what people play professionally (or at least
> is
> >not what is most famous for being played professionally), so it is going
> to
> >stand alone for a little while; it hasn't been the main focus yet. I
> >understand that it technically has features such as: very tiny point
> >differences; mostly being tactical. I don't think or have reason to
> believe
> >that that makes it somehow immune.
> >
> >What concerns me is pseudo-technical explanations for why it's harder to
> >beat humans at 9x9 than at 19x19. Saying that it's harder at 9x9 seems
> like
> >an excuse to explain (or hopefully justify) how the game is still in the
> >hands of humans. This feels very strongly like a justification for how "go
> >is still really hard for computers". Which, I suppose, we can break down
> >into lots of little subcases and worry about. The tiny point difference
> >issue is interesting; it means that things need to be super tight (less
> >room for sloppy play). Checkers also has this feature.
> >
> >The reality, in my unjustified opinion, is that this will be a solved
> >problem once it has obtained enough focus.
>
> I'm suspecious.  The value network (VN) is not enough for
> 9x9 because VN can't approximate value functions at enough
> detail.  This is also a problem on 19x19 but the advantages
> VN gives at silent positions is big enough (actually a few
> points) to beat top level human players.  I believe another
> idea is necessary for 9x9.
> #One possible (?) simple solution: if the inference speed of
> the policy network gets 100 or more times faster then we can
> use PN directly in rollouts.  This may make VN useless.
>
> Go is still hard for both human and computers :).
>
> Hideki
>
> >s.
> >
> >
> >On Fri, Feb 23, 2018 at 6:12 PM, Hideki Kato 
> wrote:
> >
> >> uurtamo .:  >> 1vhk7t...@mail.gmail.com>:
> >> >Slow down there, hombre.
> >> >
> >> >There's no secret sauce to 9x9 other than that it isn't the current
> focus
> >> >of people.
> >> >
> >> >Just like 7x7 isn't immune.
> >> >
> >> >A computer program for 9x9, funded, backed by halfway serious people,
> and
> >> >focused on the task, will *destroy* human opponents at any time it
> needs
> >> to.
> >>
> >> Why do you think (or believe) so?  I'd like to say there
> >> is no evidence so far.
> >>
> >> >If you believe that there is a special reason that 9x9 is harder than
> >> >19x19, then I'm super interested to hear that. But it's not harder for
> >> >computers. It's just not what people have been focusing on.
> >>
> >> 9x9 is not harder than 19x19 as a game.  However:  (1) Value
> >> networks, the key components to beat human on 19x19, work
> >> fine only on static positions but 9x9 has almost no such
> >> positions.   (2) Humans can play much better on 9x9
> >> than 19x19.  Top level professionals can read-out at near
> >> end of the middle stage of a game in less than 30 min with
> >> one point accuracy of the score, for example.
> >>
> >> Humans are not good at global evaluation of larger boards so
> >> bots can beat top professionals on 19x19 but this does not
> >> apply 9x9.  The size of the board is important because
> >> value networks are not universal, ie, approximate the
> >> value function not so presicely, mainly due to
> >> the number of training data is limited in practice (up to
> >> 10^8 while the number of possible input positions is greater
> >> than, at least, 10^20).  One more reason, there are no
> >> algorithm to solve double ko. This is not so big problem on
> >> 19x19 but 9x9.
> >>
> >> Best, Hideki
> >>
> >> >s.
> >> >
> >> >On Feb 23, 2018 4:49 PM, "Hideki Kato"  wrote:
> >> >
> >> >> That's not the point, Petri.  9x9 has almost no "silent"
> >> >> or "static" positons which value networks superb humans.
> >> >> On 9x9 boards, Kos, especially dou

Re: [Computer-go] 9x9 is last frontier?

[uurtamo .]

I didn't mean to suggest that I can or will solve this problem tomorrow.

What I meant to say is that it is clearly obvious that 9x9 is not immune to
being destroyed -- it's not what people play professionally (or at least is
not what is most famous for being played professionally), so it is going to
stand alone for a little while; it hasn't been the main focus yet. I
understand that it technically has features such as: very tiny point
differences; mostly being tactical. I don't think or have reason to believe
that that makes it somehow immune.

What concerns me is pseudo-technical explanations for why it's harder to
beat humans at 9x9 than at 19x19. Saying that it's harder at 9x9 seems like
an excuse to explain (or hopefully justify) how the game is still in the
hands of humans. This feels very strongly like a justification for how "go
is still really hard for computers". Which, I suppose, we can break down
into lots of little subcases and worry about. The tiny point difference
issue is interesting; it means that things need to be super tight (less
room for sloppy play). Checkers also has this feature.

The reality, in my unjustified opinion, is that this will be a solved
problem once it has obtained enough focus.

s.


On Fri, Feb 23, 2018 at 6:12 PM, Hideki Kato  wrote:

> uurtamo .:  1vhk7t...@mail.gmail.com>:
> >Slow down there, hombre.
> >
> >There's no secret sauce to 9x9 other than that it isn't the current focus
> >of people.
> >
> >Just like 7x7 isn't immune.
> >
> >A computer program for 9x9, funded, backed by halfway serious people, and
> >focused on the task, will *destroy* human opponents at any time it needs
> to.
>
> Why do you think (or believe) so?  I'd like to say there
> is no evidence so far.
>
> >If you believe that there is a special reason that 9x9 is harder than
> >19x19, then I'm super interested to hear that. But it's not harder for
> >computers. It's just not what people have been focusing on.
>
> 9x9 is not harder than 19x19 as a game.  However:  (1) Value
> networks, the key components to beat human on 19x19, work
> fine only on static positions but 9x9 has almost no such
> positions.   (2) Humans can play much better on 9x9
> than 19x19.  Top level professionals can read-out at near
> end of the middle stage of a game in less than 30 min with
> one point accuracy of the score, for example.
>
> Humans are not good at global evaluation of larger boards so
> bots can beat top professionals on 19x19 but this does not
> apply 9x9.  The size of the board is important because
> value networks are not universal, ie, approximate the
> value function not so presicely, mainly due to
> the number of training data is limited in practice (up to
> 10^8 while the number of possible input positions is greater
> than, at least, 10^20).  One more reason, there are no
> algorithm to solve double ko. This is not so big problem on
> 19x19 but 9x9.
>
> Best, Hideki
>
> >s.
> >
> >On Feb 23, 2018 4:49 PM, "Hideki Kato"  wrote:
> >
> >> That's not the point, Petri.  9x9 has almost no "silent"
> >> or "static" positons which value networks superb humans.
> >> On 9x9 boards, Kos, especially double Kos and two step Kos
> >> are important but MCTS still works worse for them, for
> >> examples.  Human professionals are much better at life&death
> >> and complex local fights which dominate small board games
> >> because they can read deterministically and deeper than
> >> current MCTS bots in standard time settings (not blitz).
> >> Also it's well known that MCTS is not good at finding narrow
> >> and deep paths to win due to "averaging".  Ohashi 6p said
> >> that he couldn't lose against statiscal algorithms after the
> >> event in 2012.
> >>
> >> Best,
> >> Hideki
> >>
> >> Petri Pitkanen:  >> 3zrby3k9kjvmzah...@mail.gmail.com>:
> >> >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 interest in 9x9.
> >> >> But it seems value network does not work well in 9x9.
> >> >> Weights_33_400 is maybe made by selfplay network. But it is 2946 in
> >9x9.
> >> &g

Re: [Computer-go] 9x9 is last frontier?

[uurtamo .]

Slow down there, hombre.

There's no secret sauce to 9x9 other than that it isn't the current focus
of people.

Just like 7x7 isn't immune.

A computer program for 9x9, funded, backed by halfway serious people, and
focused on the task, will *destroy* human opponents at any time it needs to.

If you believe that there is a special reason that 9x9 is harder than
19x19, then I'm super interested to hear that. But it's not harder for
computers. It's just not what people have been focusing on.

s.

On Feb 23, 2018 4:49 PM, "Hideki Kato"  wrote:

> That's not the point, Petri.  9x9 has almost no "silent"
> or "static" positons which value networks superb humans.
> On 9x9 boards, Kos, especially double Kos and two step Kos
> are important but MCTS still works worse for them, for
> examples.  Human professionals are much better at life&death
> and complex local fights which dominate small board games
> because they can read deterministically and deeper than
> current MCTS bots in standard time settings (not blitz).
> Also it's well known that MCTS is not good at finding narrow
> and deep paths to win due to "averaging".  Ohashi 6p said
> that he couldn't lose against statiscal algorithms after the
> event in 2012.
>
> Best,
> Hideki
>
> Petri Pitkanen:  3zrby3k9kjvmzah...@mail.gmail.com>:
> >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 interest in 9x9.
> >> But it seems value network does not work well in 9x9.
> >> Weights_33_400 is maybe made by selfplay network. But it is 2946 in 9x9.
> >> Weights_31_3200 is 4069 in 19x19 though.
> >>
> >> In year 2012, Zen played 6 games against 3 Japanese Pros, and lost by
> 0-6.
> >> And it seems Zen's 9x9 strength does not change big even now.
> >> http://computer-go.org/pipermail/computer-go/2012-November/005556.html
> >>
> >> I feel there is still enough chance that human can beat best program in
> >> 9x9.
> >>
> >> Thanks,
> >> Hiroshi Yamashita
> >>
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> > inline file
> >___
> >Computer-go mailing list
> >Computer-go@computer-go.org
> >http://computer-go.org/mailman/listinfo/computer-go
> --
> Hideki Kato 
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Project Leela Zero

[uurtamo .]

4dan?

On Jan 9, 2018 3:26 PM, "mic"  wrote:

> Thank you very much.
>
> It will play on DGS as LeelaZero19 (without GPU support). I will start
> this night with an unrated test game against FuegoBot, one minute thinking
> time each. Then I will give it a rank, so it can get stronger by the time.
>
> Leela 0.11.0 (on DGS as LeelaBot19) has meanwhile reached a weak 4 dan
> rank, also without GPU support.
>
> -Michael.
>
> ---
> Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
> https://www.avast.com/antivirus
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] mcts and tactics

[uurtamo .]

You guys are killing me.

Let's do what the space science guys did;

Parallelize via slow computation. If you need me to handle errors, I can do
ecc's. I know about how to correct for errors.

Why are we all trying to find compute power independently? Let's just add
it up. There's no real money here, let's go for it.

Break it down

s.

On Dec 19, 2017 4:26 PM, "Stephan K"  wrote:

> 2017-12-20 0:26 UTC+01:00, Dan :
> > Hello all,
> >
> > It is known that MCTS's week point is tactics. How is AlphaZero able to
> > resolve Go tactics such as ladders efficiently? If I recall correctly
> many
> > people were asking the same question during the Lee Sedo match -- and it
> > seemed it didn't have any problem with ladders and such.
>
> Note that the input to the neural networks in the version that played
> against Lee Sedol had a lot of handcrafted features, including
> information about ladders. See "extended data table 2", page 11 of the
> Nature article. You can imagine that as watching the go board through
> goggles that put a flag on each intersection that would result in a
> successful ladder capture, and another flag on each intersection that
> would result in a successful ladder escape.
>
> (It also means that you only need to read one move ahead to see
> whether a move is a successful ladder breaker or not.)
>
> Of course, your question still stands for the Zero versions.
>
> Here is the table :
>
> Feature # of planes Description
>
> Stone colour3   Player stone /
> opponent stone / empty
> Ones1   A constant plane
> filled with 1
> Turns since 8   How many turns
> since a move was played
> Liberties   8   Number of
> liberties (empty adjacent points)
> Capture size8   How many opponent
> stones would be captured
> Self-atari size 8   How many of own
> stones would be captured
> Liberties after move8   Number of
> liberties after this move is played
> Ladder capture  1   Whether a move at this
> point is a successful ladder capture
> Ladder escape   1   Whether a move at
> this point is a successful ladder escape
> Sensibleness1   Whether a move is
> legal and does not fill its own eyes
> Zeros   1   A constant plane
> filled with 0
>
> Player color1   Whether current
> player is black
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Learning related stuff

[uurtamo .]

It's nearly comic to imagine a player at 1,1 trying to figure things out.

It's not a diss on you; I honestly want for people to relax, take a minute,
and treat badmouthing the alpha go team's ideas as a secondary
consideration. They did good work. Probably arguing about the essentials
won't prove that they're stupid in any way. So let's learn, move forward,
and have no bad words about their ridiculously well-funded effort.

Recreating their work at a smaller scale would be awesome.

s.

On Nov 29, 2017 4:33 PM, "Eric Boesch"  wrote:

> Could you be reading too much into my comment? AlphaGo Zero is an amazing
> achievement, and I might guess its programmers will succeed in applying
> their methods to other fields. Nonetheless, I thought it was interesting,
> and it would appear the programmers did too, that before improving to
> superhuman level, AlphaGo was temporarily stuck in a rut of playing
> literally the worst first move on the board (excluding pass). That doesn't
> mean I think I could do better.
>
>
> On Tue, Nov 28, 2017 at 4:50 AM, uurtamo .  wrote:
>
>> This is starting to feel like asking along the lines of, "how can I
>> explain this to myself or improve on what's already been done in a way that
>> will make this whole process work faster on my hardware".
>>
>> It really doesn't look like there are a bunch of obvious shortcuts.
>> That's the whole point of decision-trees imposed by humans for 20+ years on
>> the game; it wasn't really better.
>>
>> Probably what would be good to convince oneself of these things would be
>> to challenge each assumption in divergent branches (suggested earlier) and
>> watch the resulting players' strength over time. Yes, this might take a
>> year or more on your hardware.
>>
>> I feel like maybe a lot of this is sour grapes; let's  please again
>> acknowledge that the hobbyists aren't there yet without trying to tear down
>> the accomplishments of others.
>>
>> s.
>>
>> On Nov 27, 2017 7:36 PM, "Eric Boesch"  wrote:
>>
>>> I imagine implementation determines whether transferred knowledge is
>>> helpful. It's like asking whether forgetting is a problem -- it often is,
>>> but evidently not for AlphaGo Zero.
>>>
>>> One crude way to encourage stability is to include an explicit or
>>> implicit age parameter that forces the program to perform smaller
>>> modifications to its state during later stages. If the parameters you copy
>>> from problem A to problem B also include that age parameter, so the network
>>> acts old even though it is faced with a new problem, then its initial
>>> exploration may be inefficient. For an MCTS based example, if a MCTS node
>>> is initialized to a 10877-6771 win/loss record based on evaluations under
>>> slightly different game rules, then with a naive implementation, even if
>>> the program discovers the right refutation under the new rules right away,
>>> it would still need to revisit that node thousands of times to convince
>>> itself the node is now probably a losing position.
>>>
>>> But unlearning bad plans in a reasonable time frame is already a feature
>>> you need from a good learning algorithm. Even AlphaGo almost fell into trap
>>> states; from their paper, it appears that it stuck with 1-1 as an opening
>>> move for much longer than you would expect from a program probably already
>>> much better than 40 kyu. Even if it's unrealistic for Go specifically, you
>>> could imagine some other game where after days of analysis, the program
>>> suddenly discovers a reliable trick that adds one point for white to every
>>> single game. The effect would be the same as your komi change -- a mature
>>> network now needs to adapt to a general shift in the final score. So the
>>> task of adapting to handle similar games may be similar to the task of
>>> adapting to analysis reversals within a single game, and improvements to
>>> one could lead to improvements to the other.
>>>
>>>
>>>
>>> On Fri, Nov 24, 2017 at 7:54 AM, Stephan K 
>>> wrote:
>>>
>>>> 2017-11-21 23:27 UTC+01:00, "Ingo Althöfer" <3-hirn-ver...@gmx.de>:
>>>> > My understanding is that the AlphaGo hardware is standing
>>>> > somewhere in London, idle and waitung for new action...
>>>> >
>>>> > Ingo.
>>>>
>>>> The announcement at
>>>&

Re: [Computer-go] Learning related stuff

[uurtamo .]

This is starting to feel like asking along the lines of, "how can I explain
this to myself or improve on what's already been done in a way that will
make this whole process work faster on my hardware".

It really doesn't look like there are a bunch of obvious shortcuts. That's
the whole point of decision-trees imposed by humans for 20+ years on the
game; it wasn't really better.

Probably what would be good to convince oneself of these things would be to
challenge each assumption in divergent branches (suggested earlier) and
watch the resulting players' strength over time. Yes, this might take a
year or more on your hardware.

I feel like maybe a lot of this is sour grapes; let's  please again
acknowledge that the hobbyists aren't there yet without trying to tear down
the accomplishments of others.

s.

On Nov 27, 2017 7:36 PM, "Eric Boesch"  wrote:

> I imagine implementation determines whether transferred knowledge is
> helpful. It's like asking whether forgetting is a problem -- it often is,
> but evidently not for AlphaGo Zero.
>
> One crude way to encourage stability is to include an explicit or implicit
> age parameter that forces the program to perform smaller modifications to
> its state during later stages. If the parameters you copy from problem A to
> problem B also include that age parameter, so the network acts old even
> though it is faced with a new problem, then its initial exploration may be
> inefficient. For an MCTS based example, if a MCTS node is initialized to a
> 10877-6771 win/loss record based on evaluations under slightly different
> game rules, then with a naive implementation, even if the program discovers
> the right refutation under the new rules right away, it would still need to
> revisit that node thousands of times to convince itself the node is now
> probably a losing position.
>
> But unlearning bad plans in a reasonable time frame is already a feature
> you need from a good learning algorithm. Even AlphaGo almost fell into trap
> states; from their paper, it appears that it stuck with 1-1 as an opening
> move for much longer than you would expect from a program probably already
> much better than 40 kyu. Even if it's unrealistic for Go specifically, you
> could imagine some other game where after days of analysis, the program
> suddenly discovers a reliable trick that adds one point for white to every
> single game. The effect would be the same as your komi change -- a mature
> network now needs to adapt to a general shift in the final score. So the
> task of adapting to handle similar games may be similar to the task of
> adapting to analysis reversals within a single game, and improvements to
> one could lead to improvements to the other.
>
>
>
> On Fri, Nov 24, 2017 at 7:54 AM, Stephan K 
> wrote:
>
>> 2017-11-21 23:27 UTC+01:00, "Ingo Althöfer" <3-hirn-ver...@gmx.de>:
>> > My understanding is that the AlphaGo hardware is standing
>> > somewhere in London, idle and waitung for new action...
>> >
>> > Ingo.
>>
>> The announcement at
>> https://deepmind.com/blog/applying-machine-learning-mammography/ seems
>> to disagree:
>>
>> "Our partners in this project wanted researchers at both DeepMind and
>> Google involved in this research so that the project could take
>> advantage of the AI expertise in both teams, as well as Google’s
>> supercomputing infrastructure - widely regarded as one of the best in
>> the world, and the same global infrastructure that powered DeepMind’s
>> victory over the world champion at the ancient game of Go."
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Is MCTS needed?

[uurtamo .]

Hideki,

This is a very nice observation.

s.


On Nov 16, 2017 12:37 PM, "Hideki Kato"  wrote:

Hi,

I strongly believe adding rollout makes Zero stronger.
They removed rollout just to say "no human knowledge".
#Though the number of past moves (16) has been tuned by
human :).

Hideki

Petr Baudis: <20171116154309.tfq5ix2hzwzci...@machine.or.cz>:
>  Hi,
>
>  when explaining AlphaGo Zero to a machine learning audience yesterday
>
>
>(https://docs.google.com/presentation/d/1VIueYgFciGr9pxiGmoQyUQ088Ca4o
uvEFDPoWpRO4oQ/view)
>
>it occurred to me that using MCTS in this setup is actually such
>a kludge!
>
>  Originally, we used MCTS because with the repeated simulations,
>we would be improving the accuracy of the arm reward estimates.  MCTS
>policies assume stationary distributions, which is violated every time
>we expand the tree, but it's an okay tradeoff if all you feed into the
>tree are rewards in the form of just Bernoulli trials.  Moreover, you
>could argue evaluations are somewhat monotonic with increasing node
>depths as you are basically just fixing a growing prefix of the MC
>simulation.
>
>  But now, we expand the nodes literally all the time, breaking the
>stationarity possibly in drastic ways.  There are no reevaluations that
>would improve your estimate.  The input isn't binary but an estimate in
>a continuous space.  Suddenly the Multi-armed Bandit analogy loses a lot
>of ground.
>
>  Therefore, can't we take the next step, and do away with MCTS?  Is
>there a theoretical viewpoint from which it still makes sense as the best
>policy improvement operator?
>
>  What would you say is the current state-of-art game tree search for
>chess?  That's a very unfamiliar world for me, to be honest all I really
>know is MCTS...
>
>--
>   Petr Baudis, Rossum
>   Run before you walk! Fly before you crawl! Keep moving forward!
>   If we fail, I'd rather fail really hugely.  -- Moist von Lipwig
>___
>Computer-go mailing list
>Computer-go@computer-go.org
>http://computer-go.org/mailman/listinfo/computer-go
--
Hideki Kato 
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo Zero self-play temperature

[uurtamo .]

It's interesting to leave unused parameters or unnecessary
parameterizations in the paper. It telegraphs what was being tried as
opposed to simply writing something more concise and leaving the reader to
wonder why and how those decisions were made.

s.

On Nov 7, 2017 10:54 PM, "Imran Hendley"  wrote:

> Great, thanks guys!
>
> On Tue, Nov 7, 2017 at 1:51 PM, Gian-Carlo Pascutto  wrote:
>
>> On 7/11/2017 19:07, Imran Hendley wrote:
>> > Am I understanding this correctly?
>>
>> Yes.
>>
>> It's possible they had in-betweens or experimented with variations at
>> some point, then settled on the simplest case. You can vary the
>> randomness if you define it as a softmax with varying temperature,
>> that's harder if you only define the policy as select best or select
>> proportionally.
>>
>> --
>> GCP
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo Zero self-play temperature

[uurtamo .]

If I understand your question correctly, "goes to 1" can happen as quickly
or slowly as you'd like. Yes?

On Nov 7, 2017 7:26 PM, "Imran Hendley"  wrote:

Hi, I might be having trouble understanding the self-play policy for
AlphaGo Zero. Can someone let me know if I'm on the right track here?

The paper states:

In each position s, an MCTS search is executed, guided by the neural
network f_θ . The
MCTS search outputs probabilities π of playing each move.


This wasn't clear at first since MCTS outputs wins and visits, but later
the paper explains further:

MCTS may be viewed as a self-play algorithm that, given neural
network parameters θ and a root position s, computes a vector of search
probabilities recommending moves to play, π =​  α_θ(s), proportional to
the exponentiated visit count for each move, π_a ∝​  N(s, a)^(1/τ) , where
τ is
a temperature parameter.


So this makes sense, but when I looked for the schedule for decaying the
temperature all I found was the following in the Self-play section of
Methods:


For the first 30 moves of each game, the temperature is set to τ = ​1; this
selects moves proportionally to their visit count in MCTS, and ensures a
diverse
set of positions are encountered. For the remainder of the game, an
infinitesimal
temperature is used, τ→​0.

This sounds like they are sampling proportional to visits for the first 30
moves since τ = ​1 makes the exponent go away, and after that they are
playing the move with the most visits, since the probability of the move
with the most visits goes to 1 and the probability of all other moves goes
to zero in the expression π(a | s_0) = N(s_0 , a)^(1/τ) / ∑ b N(s_0 ,
b)^(1/τ) as τ goes to 0 from the right.

Am I understanding this correctly? I am confused because it seems a little
convoluted to define this simple policy in terms of a temperature. When
they mentioned temperature I was expecting something that slowly decays
over time rather than only taking two trivial values.

Thanks!


___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] NiceGoZero games during learning

[uurtamo .]

I think it'd be quite interesting to at least keep track of the winrate
over the 4d version until then (although I recognize it will be zero for
some time). Maybe when it wins one?

s.

On Nov 6, 2017 6:00 PM, "Detlef Schmicker"  wrote:

> Not in this weak state of the learned net. I measure with a net trained
> from 4d+ kgs games right now on CGOS (NG-learn-ref).
>
> This should be the line, which could be beaten by Zero after enough
> learning. If I manage to beat this version (I check every learning cycle
> 10 games against this version) than I will probably also measure the
> strength of this, but I think this will take some weeks:)
>
>
> Am 06.11.2017 um 17:05 schrieb uurtamo .:
> > Detlef,
> >
> > I misunderstand your last sentence. Do you mean that eventually you'll
> put
> > a subset of functioning nets on CGOS to measure how quickly their
> strength
> > is improving?
> >
> > s.
> >
> > On Nov 6, 2017 4:54 PM, "Detlef Schmicker"  wrote:
> >
> >> I thought it might be fun to have some games in early stage of learning
> >> from nearly Zero knowledge.
> >>
> >> I did not turn off the (relatively weak) playouts and mix them with 30%
> >> into the result from the value network. I started at an initial random
> >> neural net (small one, about 4ms on GTX970) and use a relatively wide
> >> search for MC (much much wider, than I do for good playing strength,
> >> unpruning about 5-6 moves) and 100 playouts expanding every 3 playouts,
> >> thus 33 network evaluations per move.
> >>
> >> Additionally I add Gaussian random numbers with a standard derivation of
> >> 0.02 to the policy network.
> >>
> >> With this setup I play 1000 games and do an reinforcement learning cycle
> >> with them. One cycle takes me about 5 hours.
> >>
> >> The first 2 days I did not archive games, than I noticed it might be fun
> >> having games from the training history: now I always archive one game
> >> per cycle.
> >>
> >>
> >> Here are some games ...
> >>
> >>
> >> http://physik.de/games_during_learning/
> >>
> >>
> >> I will probably add some more games, if I have them and will try to
> >> measure, how strong the bot is with exactly this (weak broad search )
> >> configuration but a pretrained net from 4d+ kgs games on CGOS...
> >>
> >>
> >> Detlef
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> >
> >
> >
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> >
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] NiceGoZero games during learning

[uurtamo .]

Detlef,

I misunderstand your last sentence. Do you mean that eventually you'll put
a subset of functioning nets on CGOS to measure how quickly their strength
is improving?

s.

On Nov 6, 2017 4:54 PM, "Detlef Schmicker"  wrote:

> I thought it might be fun to have some games in early stage of learning
> from nearly Zero knowledge.
>
> I did not turn off the (relatively weak) playouts and mix them with 30%
> into the result from the value network. I started at an initial random
> neural net (small one, about 4ms on GTX970) and use a relatively wide
> search for MC (much much wider, than I do for good playing strength,
> unpruning about 5-6 moves) and 100 playouts expanding every 3 playouts,
> thus 33 network evaluations per move.
>
> Additionally I add Gaussian random numbers with a standard derivation of
> 0.02 to the policy network.
>
> With this setup I play 1000 games and do an reinforcement learning cycle
> with them. One cycle takes me about 5 hours.
>
> The first 2 days I did not archive games, than I noticed it might be fun
> having games from the training history: now I always archive one game
> per cycle.
>
>
> Here are some games ...
>
>
> http://physik.de/games_during_learning/
>
>
> I will probably add some more games, if I have them and will try to
> measure, how strong the bot is with exactly this (weak broad search )
> configuration but a pretrained net from 4d+ kgs games on CGOS...
>
>
> Detlef
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo Zero SGF - Free Use or Copyright?

[uurtamo .]

By way of comparison.

It would be ludicrous to ask a world champion chess player to explain their
strategy in a "programmable" way. it would certainly result in a player
much worse than the best computer player, if it were to be coded up, even
if you spent 40 years decoding intuition, etc, and got it exactly correct.

Why do I say this? Because the best human player will lose > 90% of the
time against the best computer player. And they understand their own
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 10:37 AM, Robert Jasiek  wrote:

> On 27.10.2017 13:58, Petri Pitkanen wrote:
>
>> doubt that your theory is any better than some competing ones.
>>
>
> For some specialised topics, it is evident that my theory is better or
> belongs to the few applicable theories (often by other amateur-player
> researchers) worth considering.
>
> For a broad sense of "covering every aspect of go theory", I ask: what
> competing theories? E.g., take verbal theory teaching by professional
> players and they say, e.g., "Follow the natural flow of the game". I have
> heard this for decades but still do not have the slightest idea what it
> might mean. It assumes meaning only if I replace it by my theory. Or they
> say: "Respect the beauty of shapes!" I have no idea what this means.
>
> A few particular professional players have reasonable theories on specific
> topics and resembling methodical approach occurring in my theories.
>
> So what competing theories do you mean?
>
> The heritage of professional shape examples? If you want to call that
> theory.
>
> As I do know people who are stronger than you and are using different
>> framework.
>>
>
> Yes, but where do they describe it? Almost all professional players I have
> asked to explain their decision-making have said that they could not
> because it would be intuition. A framework that is NOT theory.
>
>
> --
> robert jasiek
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Source code (Was: Reducing network size? (Was: AlphaGo Zero))

[uurtamo .]

I ask because there are (nearly) bus-speed networks that could make
multiple evaluation quick, especially if the various versions didn't differ
by more than a fixed fraction of nodes.

s.

On Oct 25, 2017 3:03 PM, uurt...@gmail.com wrote:

Does the self-play step use the most recent network for each move?

On Oct 25, 2017 2:23 PM, "Gian-Carlo Pascutto"  wrote:

On 25-10-17 17:57, Xavier Combelle wrote:
> Is there some way to distribute learning of a neural network ?

Learning as in training the DCNN, not really unless there are high
bandwidth links between the machines (AFAIK - unless the state of the
art changed?).

Learning as in generating self-play games: yes. Especially if you update
the network only every 25 000 games.

My understanding is that this task is much more bottlenecked on game
generation than on DCNN training, until you get quite a bit of machines
that generate games.

--
GCP
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Source code (Was: Reducing network size? (Was: AlphaGo Zero))

[uurtamo .]

Does the self-play step use the most recent network for each move?

On Oct 25, 2017 2:23 PM, "Gian-Carlo Pascutto"  wrote:

> On 25-10-17 17:57, Xavier Combelle wrote:
> > Is there some way to distribute learning of a neural network ?
>
> Learning as in training the DCNN, not really unless there are high
> bandwidth links between the machines (AFAIK - unless the state of the
> art changed?).
>
> Learning as in generating self-play games: yes. Especially if you update
> the network only every 25 000 games.
>
> My understanding is that this task is much more bottlenecked on game
> generation than on DCNN training, until you get quite a bit of machines
> that generate games.
>
> --
> GCP
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo Zero SGF - Free Use or Copyright?

[uurtamo .]

We're suffering under the burden of so much success from other methods that
​it's hard for many people to imagine that anything else is worth
considering.

Of course this is not true.

Tromp's enumerations are particularly enjoyable for me.

Human-built decision trees have been so unsuccessful, compared with
machine-learned models, for around 25 years, that only a few tiny wisps of
academia are interested in them in a serious way that industry can and
should take seriously.

Some control-system methods, some ILP, some NLP, etc., are all successful
counterexamples, in many cases in the field of logistics, transportation,
etc. Complicated games such as go have pretty much not fallen due to these
methods.

(As a coworker of mine said recently, "It's probably going to be okay to
hard-code the rule for the self-driving car not to hit pedestrians; there's
no need to train with lots of examples of hitting pedestrians to train your
algorithm".)

They (analytically exact methods) are still interesting to study from a
game-theoretic persepective, mathematically. There are exact solvers for
all kinds of specialized problems.

Problems with more than a few variables can very easily lead to many or
most cases not being exactly (analytically) soluble. That's why all of
these probabilistic approximation methods are so successful. They don't
have to be exactly right. It's easing the constraint most people care least
about (exact certitude of a win or success locally rather than an extremely
high probability of a win or success locally).

Asking the "high probability of success" guys to explain why their method
works is a particularly galling (and trite) way of messing with them. They
can just point to the results. The reason is that they don't know. And it's
going to be a very, very long time before they do.

At a fundamental level, probabilistic methods seem to be (some
theoreticians believe) more powerful than non-probabilistic methods for
relatively hard (as opposed to very very hard) problems. This is nicely
encoded by computational complexity theorists as the question BPP = P ?

steve



On Tue, Oct 24, 2017 at 1:41 PM, Robert Jasiek  wrote:

> On 24.10.2017 20:19, Xavier Combelle wrote:
>
>> totally unrelated
>>
>
> No, because a) software must also be evaluated and can by go theory and b)
> software can be built on exact go theory. That currently (b) is unpopular
> does not mean unrelated.
>
> --
> robert jasiek
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Alphago Zero special circumstances

[uurtamo .]

It will be interesting to realize that those specialized positions
(thousand-year-ko, bent 4) are actually a microscopic issue in game-winning.

s.


On Mon, Oct 23, 2017 at 2:10 PM, uurtamo .  wrote:

> We can all "wonder" such things unless we are not too busy to build some
> code to filter out such positions and see what actually happened in the
> self-play games opened up to everyone to see.
>
> s.
>
>
> On Mon, Oct 23, 2017 at 11:29 AM, Dave Dyer  wrote:
>
>>
>> I wonder how alphago-0 treats the menagerie of special positions, such as
>> bent 4 in the corner, thousand year ko, rotating ko, etc.
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Alphago Zero special circumstances

[uurtamo .]

We can all "wonder" such things unless we are not too busy to build some
code to filter out such positions and see what actually happened in the
self-play games opened up to everyone to see.

s.


On Mon, Oct 23, 2017 at 11:29 AM, Dave Dyer  wrote:

>
> I wonder how alphago-0 treats the menagerie of special positions, such as
> bent 4 in the corner, thousand year ko, rotating ko, etc.
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo Zero

[uurtamo .]

This sounds like a nice idea that is a misguided project.

Keep in mind the number of weights to change, and the fact that "one factor
at a time" testing will tell you nearly nothing about the overall dynamics
in a system of tens of thousands of dimensions. So you're going to need to
do something like really careful experimental design across many dimensions
simultaneously (node weights) and several million experiments -- each of
which will require hundreds if not tens of thousands of games to find the
result of the change. Worse, there are probably tens of millions of neural
nets of this size that will perform equally well (isomorphisms plus minor
weight changes). So many changes will result in no change or a completely
useless game model.

"modeling through human knowledge" neural nets doesn't sound like a
sensible goal -- it sounds more like a need to understand a topic in a
language not equipped for it without a simultaneous desire to understand a
topic under its own fundamental requirements in its own language.

Or you could build a machine-learning model to try to model those
changes except that you'd end up where you started, roughly. Another
black box and another frustrated human.

Just accept that something awesome happened and that studying those things
that make it work well are more interesting than translating coefficients
into a bad understanding for people.

I'm sorry that this NN can't teach anyone how to be a better player through
anything other than kicking their ass, but it wasn't built for that.

s.


On Fri, Oct 20, 2017 at 8:24 AM, Robert Jasiek  wrote:

> On 20.10.2017 15:07, adrian.b.rob...@gmail.com wrote:
>
>> 1) Where is the semantic translation of the neural net to human theory
>>> knowledge?
>>>
>> As far as (1), if we could do it, it would mean we could relate the
>> structures embedded in the net's weight patterns to some other domain --
>>
>
> The other domain can be "human go theory". It has various forms, from
> informal via textbook to mathematically proven. Sure, it is also incomplete
> but it can cope with additions.
>
> The neural net's weights and whatnot are given. This raw data can be
> deciphered in principle. By humans, algorithms or a combination.
>
> You do not know where to start? Why, that is easy: test! Modify ONE weight
> and study its effect on ONE aspect of human go theory, such as the
> occurrance (frequency) of independent life. No effect? Increase the
> modification, test a different weight, test a subset of adjacent weights
> etc. It has been possible to study semantics of parts of DNA, e.g., from
> differences related to illnesses. Modifications on the weights is like
> creating causes for illnesses (or improved health).
>
> There is no "we cannot do it", but maybe there is too much required effort
> for it to be financially worthwhile for the "too specialised" case of Go?
> As I say, a mathematical proof of a complete solution of Go will occur
> before AI playing perfectly;)
>
> So far neural
>> nets have been trained and applied within single domains, and any
>> "generalization" means within that domain.
>>
>
> Yes.
>
> --
> robert jasiek
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Deep Blue the end, AlphaGo the beginning?

[uurtamo .]

Gian-Carlo,

I only ask, not to be snippy or impolite, but because I have just exactly
enough knowledge to be dangerous enough to have no freaking idea what I'm
talking about wrt chess research, and by way of introduction, let me say
that I've seen some people talk about (and a coworker at my former
university worked with) strong chess programs and I've done some analysis
with them. I think of them generally as black boxes whose strength gets
more and more complicated to measure since they can only essentially play
themselves anymore in an interesting way. Eventually I imagine it will take
more analysis on our part to understand their games then they are going to
give us. Which I'm fine with.

But.

They run on laptops. A program that could crush a grandmaster will run on
my laptop. That's an assertion I can't prove, but I'm asking you to verify
it or suggest otherwise.

Now the situation with go is different.

Perhaps it's that the underlying problem is harder. But "those old methods"
wouldn't work on this problem. I only mean that in the sense that the exact
code for chess, working with the rules of go, adapated using some
first-pass half-assed idea of what that means, would fail horribly.
Probably both because 64 << 169 and because queen >> 1 stone and for god
only knows how many other reasons.

So let's first get out of the way that this was probably a much harder
problem (the go problem).

I agree that the sharp definition of "machine learning", "statistics",
"AI", "blah blah blah" don't really matter toward the idea of  "computer
game players", etc.

But if we do agree that the problem itself is fundamentally harder, (which
I believe it is) and we don't want to ascribe its solution simply to
hardware (which people tried to do with big blue), then we should
acknowledge that it required more innovation.

I do agree, and hope that you do, that this innovation is all part of a
continuum of innovation that is super exciting to understand.

Thanks,

steve


​

On Fri, Aug 18, 2017 at 1:31 PM, Gian-Carlo Pascutto  wrote:

> On 18-08-17 16:56, Petr Baudis wrote:
> >> Uh, what was the argument again?
> >
> >   Well, unrelated to what you wrote :-) - that Deep Blue implemented
> > existing methods in a cool application, while AlphaGo introduced
> > some very new methods (perhaps not entirely fundamentally, but still
> > definitely a ground-breaking work).
>
> I just fundamentally disagree with this characterization, which I think
> is grossly unfair to the Chiptest/Deep Thought/Deep Blue lineage.
> Remember there were 12 years in-between those programs.
>
> They did not just...re-implement the same "existing methods" over and
> over again all that time. Implementation details and exact workings are
> very important [1]. I imagine the main reason this false distinction
> (i.e. the "artificial difference" from my original post) is being made
> is, IMHO, that you're all aware of the fine nuances of how AlphaGo DCNN
> usage (for example) differs compared to previous efforts, but you're not
> aware of the same nuances in Chiptest and successors etc.
>
> [1] As is speed, another dirty word in AI circles that is nevertheless
> damn important for practical performance.
>
> --
> GCP
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Alphago and solving Go

[uurtamo .]

Why do you think that there is a 3 in the denominator?

On Aug 9, 2017 2:29 PM, "Marc Landgraf"  wrote:

> I don't mind your terminology, in fact I feel like it is a good way to
> distinguish the two different things. It is just that I considiered one
> thing wrongly used instead of the other for the discussion here.
>
> But if we go with the link you are suggesting here:
> Shouldnt that number at most be 722^#positions? Since adding a black or a
> white stone is something fundamentally different?
>
> 2017-08-09 20:50 GMT+02:00 John Tromp :
>
>> > And what is the connection between the number of "positions" and the
>> number
>> > of games
>>
>> The number of games is at most 361^#positions.
>>
>> > or even solving games? In the game trees we do not care about
>> > positions, but about situations.
>>
>> We care about lots of things, including intersections, stones,
>> liberties, strings, positions, sets of previous positions.
>>
>> > I'm actually surprised that this "absurd" to you...
>>
>> I said that referring to a board configuration together with the set
>> of all previously occurring board configurations (and turn to move) as
>> "position" is absurd.
>> We need a simple word to denote a board configuration, and "position" fits
>> that requirement. A good word for all the relevant historical
>> information leading up to a position is "situation".
>>
>> -John
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Alphago and solving Go

[uurtamo .]

It's trivial, dude.

On Aug 9, 2017 8:35 AM, "Marc Landgraf"  wrote:

> Under which ruleset is the 3^(n*n) a trivial upper bound for the number of
> legal positions?
> I'm sure there are rulesets, under which this bonds holds, but I doubt
> that this can be considered trivial.
>
> Under the in computer go more common rulesets this upper bound is simply
> wrong. Unless we talk about simply the visual aspect, but then this has
> absolutely nothing to do with the discussion abour solving games.
>
> 2017-08-09 14:34 GMT+02:00 Gunnar Farnebäck :
>
>> Except 361! (~10^768) couldn't plausibly be an estimate of the number of
>> legal positions, since ignoring the rules in that case gives the trivial
>> upper bound of 3^361 (~10^172).
>>
>> More likely it is a very, very bad attempt at estimating the number of
>> games. Even with the extremely unsharp bound given in
>> https://tromp.github.io/go/gostate.pdf
>>
>> 10^(10^48) < number of games < 10^(10^171)
>>
>> the 361! estimate comes nowhere close to that interval.
>>
>> /Gunnar
>>
>> On 08/07/2017 04:14 AM, David Doshay wrote:
>>
>>> Yes, that zeroth order number (the one you get to without any thinking
>>> about how the game’s rules affect the calculation) is outdated since early
>>> last year when this result gave us the exact number of legal board
>>> positions:
>>>
>>> https://tromp.github.io/go/legal.html
>>>
>>> So, a complete game tree for 19x19 Go would contain about 2.08 * 10^170
>>> unique nodes (see the paper for all 171 digits) but some number of
>>> duplicates of those nodes for the different paths to each legal position.
>>>
>>> In an unfortunate bit of timing, it seems that many people missed this
>>> result because of the Alpha Go news.
>>>
>>> Cheers,
>>> David G Doshay
>>>
>>> ddos...@mac.com 
>>>
>>>
>>>
>>>
>>>
>>> On 6, Aug 2017, at 3:17 PM, Gunnar Farnebäck >>> > wrote:

 On 08/06/2017 04:39 PM, Vincent Richard wrote:

> No, simply because there are way to many possibilities in the game,
> roughly (19x19)!
>

 Can we lay this particular number to rest? Not that "possibilities in
 the game" is very well defined (what does it even mean?) but the number of
 permutations of 19x19 points has no meaningful connection to the game of go
 at all, not even "roughly".

 /Gunnar
 ___
 Computer-go mailing list
 Computer-go@computer-go.org 
 http://computer-go.org/mailman/listinfo/computer-go

>>>
>>>
>>>
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] DeepMind's Victory over Ke Jie

[uurtamo .]

1) triumphed over a particular boardgame, perhaps.

2) this has been true since the early days of Monte Carlo go bots -- it
tends to result in a higher winning probability when the focus is on...
winning probability. People have tried to make bots that focus on points
instead, but trying to win by more than komi usually results in a weaker
bot. Others can detail their experiments for you and have here at length.

s.


On Jun 7, 2017 1:21 PM, "Cai Gengyang"  wrote:

> Hi guys,
>
> Just a couple of questions :
>
> 1) Is it true that DeepMind's comprehensive victory over Ke Jie means that
> essentially it is proven to be true that AI has definitely triumphed over
> humanity ?
>
> 2) Also, I read that AG's style is "conservative" -- i.e. it almost always
> prefers the higher chance of winning by a small number of points compared
> to the smaller chance of winning by a large number of points ?
>
> Thanks alot
>
> GengYang
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Zen lost to Mi Yu Ting

[uurtamo .]

I guess that 1 point in such a game matters to the evaluation function.
Pretty fascinating. Can you not train for the two different rulesets and
just pick which at the beginning? Ignoring Chinese versus Japanese, just
training on komi? Or is the problem of Japanese rules the whole issue? (I.e
not komi)?

On Tuesday, March 21, 2017, Hideki Kato  wrote:

> The value network has been trained with Chinese rules and 7.5
> pts komi.  Using this for Japanese and 6.5, there will be some
> error in close games.  We knew this issue and thought such
> chances would be so small that postponed correcting (not so
> easy).
>
> Best,
> Hideki
>
> Pawe  Morawiecki: <
> caksbshpvd34hvjt-b+x73rdpg5-4wsxoezykbheslprewci...@mail.gmail.com
> >:
> >Hi,
> >
> >After an interesting game DeepZen lost to Mi Yu Ting.
> >Here you can replay the complete game:
> >http://duiyi.sina.com.cn/gibo_new/live/viewer.asp?sno=13
> >
> >According to pro experts, Zen fought really well, but it seems there is
> >still some issue how Zen (mis)evaluates its chances. At one point it
> showed
> >84% chance of winning (in the endgame), whereas it was already quite clear
> >Zen is little behind (2-3 points).
> >
> >Regards,
> >Pawel
> > inline file
> >___
> >Computer-go mailing list
> >Computer-go@computer-go.org 
> >http://computer-go.org/mailman/listinfo/computer-go
> --
> Hideki Kato 
> ___
> Computer-go mailing list
> Computer-go@computer-go.org 
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Zen19K2 is strongest player on KGS

[uurtamo .]

This is really good to hear.

3 stones is totally reasonable.

s.

On Oct 4, 2016 8:02 PM, "Hiroshi Yamashita"  wrote:

> Hi,
>
> Zen19K2 is strongest player on KGS.
> http://www.gokgs.com/top100.jsp
> Oops, another player is top now. But anyway nearly top.
>
> Zen19K2 is maybe 10.3d from graph.
> http://www.gokgs.com/graphPage.jsp?user=Zen19K2
>
> Zen19K2's information is -
> Computer program Zen running on KURISU server provided by DWANGO.
>
> CPU: Xeon E5-2623 v3 x2
> GPU: GeForce GTX TITAN X x4
>
> The number of handicap stones is limited to 3 or less.
> -
>
> Congratulations for graduation from KGS, Zen!
> I think Zen19K2 strength is similar to 2015/10 AlphaGo that beated Fan Hui
> 2p, 5-0.
>
> Thanks,
> Hiroshi Yamashita
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] DarkForest is open-source now.

[uurtamo]

GPL is rough
On Jun 10, 2016 2:02 PM, "Xavier Combelle" 
wrote:

> for me it's clearly GPL violation
>
> 2016-06-10 22:17 GMT+02:00 Darren Cook :
>
>> >> At 5d KGS, is this the world's strongest MIT/BSD licensed program? ...
>> >> actually, is there any other MIT/BSD go program out there? (I thought
>> >> Pachi was, but it is GPLv2)
>> >
>> > Huh, that's interesting, because Darkforest seems to have copy-pasted
>> > the pachi playout policy:
>> >
>> >
>> https://github.com/facebookresearch/darkforestGo/blob/master/board/pattern.c#L36
>> >
>> > https://github.com/pasky/pachi/blob/master/playout/moggy.c#L101
>>
>> Uh-oh. Though it does say "inspired by" at the top, and also that it is
>> not used by the main engine:
>>
>> // This file is inspired by Pachi's engine
>> //   (https://github.com/pasky/pachi).
>> // The main DarkForest engine (when specified
>> //   with `--playout_policy v2`) does not depend on it.
>> // However, the simple policy opened with
>> //   `--playout_policy simple` will use this library.
>>
>>
>> Darren
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] DarkForest is open-source now.

[uurtamo]

Compiler no workie? ;)

s.
On Jun 10, 2016 11:15 AM, "Dave Dyer"  wrote:

>
> Now if someone would post a binary that would just run on suitable
> hardware.
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Machine for Deep Neural Net training

[uurtamo .]

That's pretty awesome.

I didn't think it'd approach the 100x barrier. That's shocking.

s.
On Apr 26, 2016 9:55 PM, "David Fotland"  wrote:

> I have my deep neural net training setup working, and it's working so well
> I
> want to share.  I already had Caffe running on my desktop machine (4 core
> i7) without a GPU, with inputs similar to AlphaGo generated by Many Faces
> into an LMDB database.  I trained a few small nets for a day each to get
> some feel for it.
>
> I bought an Alienware Area 51 from Dell, with two GTX 980 TI GPUs, 16 GB of
> memory, and 2 TB of disk.  I set it up to dual boot Ubuntu 14.04, which
> made
> it trivial to get the latest caffe up and running with CUDNN.  2 TB of disk
> is not enough.  I'll have to add another drive.
>
> I expected something like 20x speedup on training, but I was shocked by
> what
> I actually got.
>
> On my desktop, the Caffe MNIST sample took 27 minutes to complete.  On the
> new machine it was 22 seconds.  73x faster.
>
> My simple network has 42 input planes, and 4 layers of 48 filters each.
> Training runs about 100x faster on the Alienware.  Training 100k Caffe
> iterations (batches) of 50 positions takes 13 minutes, rather than almost a
> full day on my desktop.
>
> David
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] OmegaGo

[uurtamo .]

Pamphlets <= treatises
On Apr 20, 2016 11:17 AM, "David Ongaro"  wrote:

> Some of "Dr. Browns" pamphlets remind me of Herman Hesse's Steppenwolf.
> But Hesse's text was much more refined and enjoyable to read, so I think
> "SPAM" is a fitting categorisation.
>
>
>
> > On 20 Apr 2016, at 00:57, Michael Markefka 
> wrote:
> >
> > Can I flag this as spam?
> >
> > On Tue, Apr 19, 2016 at 11:23 PM, djhbrown .  wrote:
> >> 6D out of the blue is no mean achievement,...  60+ years ago, the
> >> market for gizmos in UK was flooded with cheap Japanese copies of
> >> European products; but whilst innovation and product quality
> >> improvement by European manufacturers faded as their fat cat owners
> >> complacently went cocacola-soaked soft,  Japanese industry, unlike its
> >> USA counterpart, was listening attentively to the wise words of
> >> W.Edwards Deming (eg [1,2]) and beginning to improve the reliability,
> >> efficiency and efficacy of its products, and by about 30 years ago,
> >> Japanese engineering was the equal or better of even German
> >> technology.
> >>
> >> Korean, Formosan and Hong Kong e-tigers followed hotfoot in Japan's
> >> footsteps, and now the same thing is happening in China, so we can
> >> expect to see a vast array of Shanghai-teenager-bedroom-produced
> >> shanghaied miniclones of Alpha, most with unimaginative copycat names
> >> like Beta, Eta, Theta, AIota etc, skulking around the corridors of the
> >> Internet, all of which will at first be cheap imitations, but sowing
> >> the seeds of in-house and inter-house R&D quality circles, so that
> >> their own descendants will before very long become to Californian IT
> >> as Japanese fuel-efficient reliable engines are to US unreliable
> >> gas-guzzlers.
> >>
> >> Watch out Google Cloud byte-guzzlers, teenage rebels with the lessons
> >> of Deming in their notebooks, who have learned from history and from
> >> the sterling modus operandi of Steve Jobs and Uncle Tom Cobley et al,
> >> are on their way up your Jacob's ladder...
> >>
> >> 1.  Charles A. Barclay (1993) Quality Strategy and TQM Policies:
> >> Empirical Evidence.
> >> MIR: Management International Review Vol. 33, Strategic Quality
> Management.
> >> 2.
> http://asq.org/learn-about-quality/total-quality-management/overview/deming-points.html
> >>
>  Anybody knows who is the author of BetaGo? It is playing with account
>  GoBeta on KGS, and is 6d.
> >> ___
> >> Computer-go mailing list
> >> Computer-go@computer-go.org
> >> http://computer-go.org/mailman/listinfo/computer-go
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] new challenge for Go programmers

[uurtamo .]

Nice
On Mar 31, 2016 7:48 AM, "Álvaro Begué"  wrote:

> A very simple-minded way of trying to identify what a particular neuron in
> the upper layers is doing is to find the 50 positions in the database that
> make it produce the highest activation values. If the neuron is in one of
> the convolutional layers, you get a full 19x19 image of activation values,
> which would let you figure out what particular local pattern it seems to be
> detecting. If the neuron is in a fully-connected layer at the end, you only
> get one overall value, but you could still try to compute the gradient of
> its activation with respect to all the inputs, and that would tell you
> something about what parts of the board led to this activation being high.
> I think this would be a fun exercise, and you'll probably be able to
> understand something about at least some of the neurons.
>
> Álvaro.
>
>
>
> On Thu, Mar 31, 2016 at 9:55 AM, Michael Markefka <
> michael.marke...@gmail.com> wrote:
>
>> Then again DNNs also manage feature extraction on unlabeled data with
>> increasing levels of abstraction towards upper layers. Perhaps one
>> could apply such a specifically trained DNN to artificial board
>> situations that emphasize specific concepts and examine the network's
>> activation, trying to map activation patterns to human Go concepts.
>>
>> Still hard work, and questionable payoff, but just wanted to pitch
>> that in as idea.
>>
>>
>> > However, if someone was to do all the dirty work setting up all the
>> > infrastructure, hunt down the training data and then financially
>> facilitate
>> > the thousands of hours of human work and the tens to hundreds of
>> thousands
>> > of hours of automated learning work, I would become substantially more
>> > interested...and think a high quality desired outcome remains a low
>> > probability.
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] "English Explanations" based on Neural Networks

[uurtamo .]

Major changes in the evaluation probability could likely have a horizon of
a few moves behind that might be interesting to more closely evaluate. With
a small window like that, a deeper/more exhaustive search might work.

s.
On Mar 31, 2016 10:21 AM, "Petr Baudis"  wrote:

> On Thu, Mar 31, 2016 at 08:51:30AM -0500, Jim O'Flaherty wrote:
> > What I was addressing was more around what Robert Jasiek is describing in
> > his joseki books and other materials he's produced. And it is exactly
> why I
> > think the "explanation of the suggested moves" requires a much deeper
> > baking into the participating ANN's (bottom up approach). And given what
> I
> > have read thus far (including your above information), I am still seeing
> > the risk extraordinarily high and the payoff exceedingly low, outside an
> > academic context.
>
>   I think we may just have a different outcome in mind.  To illustrate
> where I think my approach could work, that could be for example
> (slightly edited):
>
> > White Q5 was played to compel Black to extend at the bottom.
> > If Black doesn’t respond, White’s pincer at K4 will be powerful.
>
> in
> https://gogameguru.com/lee-sedol-defeats-alphago-masterful-comeback-game-4/
>
>
>   Sure, it seems a bit outrageous, and for initial attempts, generating
> utterances like
>
> > White 126 was a very big move which helped to ensure White’s advantage.
>
> is perhaps more realistic (though many of these sentences are a bit
> of truisms and not terribly informative).  But I'm quite convinced that
> even the first example is completely plausible.
>
>   (But I'm *not* talking about generating pages of diagrams that
> describe an opening position in detail.  That's to ponder when we
> get the simpler things right.)
>
> --
> Petr Baudis
> If you have good ideas, good data and fast computers,
> you can do almost anything. -- Geoffrey Hinton
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] new challenge for Go programmers

[uurtamo .]

Fair enough
On Mar 30, 2016 5:20 PM, "Brian Sheppard"  wrote:

> This is out of line, IMO. Djhbrown asked a sensible question that has
> valuable intentions. I would like to see responsible, thoughtful, and
> constructive replies.
>
>
>
> *From:* Computer-go [mailto:computer-go-boun...@computer-go.org] *On
> Behalf Of *uurtamo .
> *Sent:* Wednesday, March 30, 2016 7:43 PM
> *To:* computer-go 
> *Subject:* Re: [Computer-go] new challenge for Go programmers
>
>
>
> He cannot possibly write code
>
> On Mar 30, 2016 4:38 PM, "Jim O'Flaherty" 
> wrote:
>
> I don't think djhbrown is a software engineer. And he seems to have the
> most fits. :)
>
>
>
> On Wed, Mar 30, 2016 at 6:37 PM, uurtamo .  wrote:
>
> This is clearly the alphago final laugh; make an email list responder to
> send programmers into fits.
>
> s.
>
> On Mar 30, 2016 4:16 PM, "djhbrown ."  wrote:
>
> thank you very much Ben for sharing the inception work, which may well
> open the door to a new avenue of AI research.  i am particularly
> impressed by one pithy statement the authors make:
>
>  "We must go deeper: Iterations"
>
> i remember as an undergrad being impressed by the expressive power of
> recursive functions, and later by the iterative quality of biological
> growth and its fractal nature.
>
> seeing animals in clouds is a bit like seeing geta in a go position;
> so maybe one way to approach the problem of chatting with a CNN might
> be to seek correlations between convolution weights and successive
> stone configurations that turn up time and time again in games.
>
> it may be that some kind of iterative procedure could do this, just as
> my iterative procedure for circumscribing a group has a recursive
> quality to its definition.
>
> all you need then is to give such a correlation a name, and you will
> be on the way to discovering a new language for talking about Go.
>
>
> On 31/03/2016, Ben  wrote:
> > It would be very interesting to see what these go playing neural
> > networks dream about [1].
> > [1]
> >
> http://googleresearch.blogspot.de/2015/06/inceptionism-going-deeper-into-neural.html
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] new challenge for Go programmers

[uurtamo .]

He cannot possibly write code
On Mar 30, 2016 4:38 PM, "Jim O'Flaherty" 
wrote:

> I don't think djhbrown is a software engineer. And he seems to have the
> most fits. :)
>
> On Wed, Mar 30, 2016 at 6:37 PM, uurtamo .  wrote:
>
>> This is clearly the alphago final laugh; make an email list responder to
>> send programmers into fits.
>>
>> s.
>> On Mar 30, 2016 4:16 PM, "djhbrown ."  wrote:
>>
>>> thank you very much Ben for sharing the inception work, which may well
>>> open the door to a new avenue of AI research.  i am particularly
>>> impressed by one pithy statement the authors make:
>>>
>>>  "We must go deeper: Iterations"
>>>
>>> i remember as an undergrad being impressed by the expressive power of
>>> recursive functions, and later by the iterative quality of biological
>>> growth and its fractal nature.
>>>
>>> seeing animals in clouds is a bit like seeing geta in a go position;
>>> so maybe one way to approach the problem of chatting with a CNN might
>>> be to seek correlations between convolution weights and successive
>>> stone configurations that turn up time and time again in games.
>>>
>>> it may be that some kind of iterative procedure could do this, just as
>>> my iterative procedure for circumscribing a group has a recursive
>>> quality to its definition.
>>>
>>> all you need then is to give such a correlation a name, and you will
>>> be on the way to discovering a new language for talking about Go.
>>>
>>>
>>> On 31/03/2016, Ben  wrote:
>>> > It would be very interesting to see what these go playing neural
>>> > networks dream about [1].
>>> > [1]
>>> >
>>> http://googleresearch.blogspot.de/2015/06/inceptionism-going-deeper-into-neural.html
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] new challenge for Go programmers

[uurtamo .]

This is clearly the alphago final laugh; make an email list responder to
send programmers into fits.

s.
On Mar 30, 2016 4:16 PM, "djhbrown ."  wrote:

> thank you very much Ben for sharing the inception work, which may well
> open the door to a new avenue of AI research.  i am particularly
> impressed by one pithy statement the authors make:
>
>  "We must go deeper: Iterations"
>
> i remember as an undergrad being impressed by the expressive power of
> recursive functions, and later by the iterative quality of biological
> growth and its fractal nature.
>
> seeing animals in clouds is a bit like seeing geta in a go position;
> so maybe one way to approach the problem of chatting with a CNN might
> be to seek correlations between convolution weights and successive
> stone configurations that turn up time and time again in games.
>
> it may be that some kind of iterative procedure could do this, just as
> my iterative procedure for circumscribing a group has a recursive
> quality to its definition.
>
> all you need then is to give such a correlation a name, and you will
> be on the way to discovering a new language for talking about Go.
>
>
> On 31/03/2016, Ben  wrote:
> > It would be very interesting to see what these go playing neural
> > networks dream about [1].
> > [1]
> >
> http://googleresearch.blogspot.de/2015/06/inceptionism-going-deeper-into-neural.html
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] new challenge for Go programmers

[uurtamo .]

Guys, please take a day.

steve
On Mar 30, 2016 1:52 PM, "Brian Sheppard"  wrote:

> Trouble is that it is very difficult to put certain concepts into
> mathematics. For instance: “well, I tried to find parameters that did a
> better job of minimizing that error function, but eventually I lost
> patience.” :-)
>
>
>
> Neural network parameters are not directly humanly understandable. They
> just happen to minimize an error function on a sample of training cases
> that might not even be representative. So you want to reason “around” the
> NN by interrogating it in some way, and trying to explain the results.
>
>
>
> If anyone wants to pursue this research, I suggest several avenues.
>
>
>
> First, you could differentiate the output with respect to each input to
> determine the aspects of the position that weigh on the result most
> heavily. Then, assuming that you can compare the scale of the inputs in
> some way, and assuming that the inputs are something that is understandable
> in the problem domain, maybe you can construct an explanation.
>
>
>
> Second, you could construct a set of hypothetical different similar
> positions, and see how those results differ. E.g., make a set of examples
> by adding a black stone and a white stone to each empty point on the board,
> or removing each existing stone from the board, and then evaluate the NN on
> those cases, then do decision-tree induction to discover patterns.
>
>
>
> Third, in theory decision trees are just as powerful as NN (in that both
> are asymptotically optimal learning systems), and it happens that decision
> trees provide humanly understandable explanations for reasoning. So maybe
> you can replace the NN with DT and have equally impressive performance, and
> pick up human understandability as a side-effect.
>
>
>
> Actually, if anyone is interested in making computer go programs that do
> not require GPUs and super-computers, then looking into DTs is advisable.
>
>
>
> Best,
>
> Brian
>
>
>
>
>
> *From:* Computer-go [mailto:computer-go-boun...@computer-go.org] *On
> Behalf Of *Jim O'Flaherty
> *Sent:* Wednesday, March 30, 2016 4:24 PM
> *To:* computer-go@computer-go.org
> *Subject:* Re: [Computer-go] new challenge for Go programmers
>
>
>
> I agree, "cannot" is too strong. But, values close enough to "extremely
> difficult as to be unlikely" is why I used it.
>
> On Mar 30, 2016 11:12 AM, "Robert Jasiek"  wrote:
>
> On 30.03.2016 16:58, Jim O'Flaherty wrote:
>
> My own study says that we cannot top down include "English explanations" of
> how the ANNs (Artificial Neural Networks, of which DCNN is just one type)
> arrive a conclusions.
>
>
> "cannot" is a strong word. I would use it only if it were proven
> mathematically.
>
> --
> robert jasiek
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Congratulations to AlphaGo (Statistical significance of results)

[uurtamo .]

Or, if it's lopsided far from 1/2, Wilson's is just as good, in my
experience.
On Mar 30, 2016 10:29 AM, "Olivier Teytaud"  wrote:

> don't use asymptotic normality with a sample size 5, use Fisher's exact
> test
>
> the p-value for the rejection of
> "P(alpha-Go wins a given game against Lee Sedol)<.5"
> might be something like 3/16
> (under the "independent coin" assumption!)
>
> this is not 0.05, but still quite an impressive result :-)
>
> with 5-0 it would have been <0.05.
>
>
>
> On Wed, Mar 30, 2016 at 6:59 PM, Ryan Hayward  wrote:
>
>> Hey Simon,
>>
>> I only now remembered:
>>
>> we actually experimented on the effect
>> of making 1 blunder (random move instead of learned/searched move)
>> in Go and Hex
>>
>> "Blunder Cost in Go and Hex"
>>
>> so this might be a starting point for your question
>> of measuring player strength by measuring
>> all move strengths...
>>
>> https://webdocs.cs.ualberta.ca/~hayward/papers/blunder.pdf
>>
>> On Wed, Mar 30, 2016 at 5:29 AM, Lucas, Simon M  wrote:
>>
>>> In my original post I put a link to
>>> the relevant section of the MacKay
>>> book that shows exactly how to calculate
>>> the probability of superiority
>>> assuming the game outcome is modelled as
>>> a biased coin toss:
>>>
>>> http://www.inference.phy.cam.ac.uk/itila/
>>>
>>>
>>> I was making the point that for this
>>>
>>> and for other outcomes of skill-based games
>>> we can do so much more (and as humans we intuitively
>>> DO do so much more) than just look at the event
>>> outcome - and maybe as a community we should do that more
>>> routinely and more quantitatively (e.g.
>>> by analysing the quality of each move / action)
>>>
>>> Best wishes,
>>>
>>>   Simon
>>>
>>>
>>>
>>> On 30/03/2016, 11:57, "Computer-go on behalf of djhbrown ." <
>>> computer-go-boun...@computer-go.org on behalf of djhbr...@gmail.com>
>>> wrote:
>>>
>>> >Simon wrote: "I was discussing the results with a colleague outside
>>> >of the Game AI area the other day when he raised
>>> >the question (which applies to nearly all sporting events,
>>> >given the small sample size involved)
>>> >of statistical significance - suggesting that on another week
>>> >the result might have been 4-1 to Lee Sedol."
>>> >
>>> >call me naive, but perhaps you could ask your colleague to calculate
>>> >the probability one of side winning 4 games out of 5, and then say
>>> >whether that is within 2 standard deviations of the norm.
>>> >
>>> >his suggestion is complete nonsense, regardless of the small sample
>>> >size.  perhaps you could ask a statistician next time.
>>> >
>>> >--
>>> >patient: "whenever i open my mouth, i get a shooting pain in my foot"
>>> >doctor: "fire!"
>>> >http://sites.google.com/site/djhbrown2/home
>>> >https://www.youtube.com/user/djhbrown
>>> >___
>>> >Computer-go mailing list
>>> >Computer-go@computer-go.org
>>> >http://computer-go.org/mailman/listinfo/computer-go
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>
>>
>>
>> --
>> Ryan B Hayward
>> Professor and Director (Outreach+Diversity)
>> Computing Science,  UAlberta
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
>
> --
> =
> Olivier Teytaud, olivier.teyt...@inria.fr, TAO, LRI, UMR 8623(CNRS -
> Univ. Paris-Sud),
> bat 490 Univ. Paris-Sud F-91405 Orsay Cedex France
> http://www.slideshare.net/teytaud
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Would a larger board (25x25) dramatically reduce AlphaGos skill?

[uurtamo .]

Ko is what makes this game difficult, from a theoretical point of view.

I suspect ko+unresolved groups is where it's at.

s.
On Mar 22, 2016 11:25 AM, "Tom M"  wrote:

> I suspect that even with a similarly large training sample for
> initialization that AlphaGo would suffer a major reduction in apparent
> skill level.  The CNN would require many more layers of convolution;
> the valuation of positions would be much more uncertain; play in the
> corner, edges, and center would all be more complicated patterns, and
> there would be far more good candidates to consider at each ply and
> rollouts would be much less stable and less accurate.
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Congratulations to AlphaGo (Statistical significance of results)

[uurtamo .]

This is somewhat moot - if any moves had been significantly and obviously
weak to any observers, the results wouldn't have been 4-1.

I.e. One bad move out of 5 games would give roughly the same strength
information as one loss out of 5 games; consider that the kibitzing was
being done in real time.

s.
On Mar 22, 2016 11:08 AM, "Jim O'Flaherty" 
wrote:

> I think you are reinforcing Simon's original point; i.e. using a more fine
> grained approach to statically approximate AlphaGo's ELO where fine grained
> is degree of vetting per move and/or a series of moves. That is a
> substantially larger sample size and each sample will have a pretty high
> degree of quality (given the vetting is being done by top level
> professionals).
> On Mar 22, 2016 1:04 PM, "Jeffrey Greenberg" 
> wrote:
>
>> Given the minimal sample size, bothering over this question won't amount
>> to much. I think the proper response is that no one thought we'd see this
>> level of play at this point in our AI efforts and point to the fact that we
>> witnessed hundreds of moves vetted by 9dan players, especially Michael
>> Redmond's, where each move was vetted. In other words "was the level of
>> play very high?" versus the question "have we beat all humans". The answer
>> is more or less, yes.
>>
>> On Tuesday, March 22, 2016, Lucas, Simon M  wrote:
>>
>>> Hi all,
>>>
>>> I was discussing the results with a colleague outside
>>> of the Game AI area the other day when he raised
>>> the question (which applies to nearly all sporting events,
>>> given the small sample size involved)
>>> of statistical significance - suggesting that on another week
>>> the result might have been 4-1 to Lee Sedol.
>>>
>>> I pointed out that in games of skill there's much more to judge than
>>> just the final
>>> outcome of each game, but wondered if anyone had any better (or worse :)
>>> arguments - or had even engaged in the same type of
>>> conversation.
>>>
>>> With AlphaGo winning 4 games to 1, from a simplistic
>>> stats point of view (with the prior assumption of a fair
>>> coin toss) you'd not be able to claim much statistical
>>> significance, yet most (me included) believe that
>>> AlphaGo is a genuinely better Go player than Lee Sedol.
>>>
>>> From a stats viewpoint you can use this approach:
>>> http://www.inference.phy.cam.ac.uk/itprnn/book.pdf
>>> (see section 3.2 on page 51)
>>>
>>> but given even priors it won't tell you much.
>>>
>>> Anyone know any good references for refuting this
>>> type of argument - the fact is of course that a game of Go
>>> is nothing like a coin toss.  Games of skill tend to base their
>>> outcomes on the result of many (in the case of Go many hundreds of)
>>> individual actions.
>>>
>>> Best wishes,
>>>
>>>   Simon
>>>
>>>
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Congratulations to AlphaGo (Statistical significance of results)

[uurtamo .]

> I'm not sure if we can say with certainty that AlphaGo is significantly
> better Go player than Lee Sedol at this point.  What we can say with
> certainty is that AlphaGo is in the same ballpark and at least roughly
> as strong as Lee Sedol.  To me, that's enough to be really huge on its
> own accord!

Agreed, and exactly what I'm telling my friends who have asked the same
question.

s.
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Congratulations to AlphaGo (Statistical significance of results)

[uurtamo .]

Simon,

There's no argument better than evidence, and no evidence available to us
other than *all* of the games that alphago has played publicly.

Among two humans, a 4-1 result wouldn't indicate any more or less than this
4-1 result, but we'd already have very strong elo-type information about
both humans because they both would have publicly played hundreds of games
to get to such a match.

I believe alphago played another match earlier in public, correct?  Then we
now have double the evidence, or a slight (50% or so) improvement in our
confidence bounds.

s.
On Mar 22, 2016 9:00 AM, "Lucas, Simon M"  wrote:

> Hi all,
>
> I was discussing the results with a colleague outside
> of the Game AI area the other day when he raised
> the question (which applies to nearly all sporting events,
> given the small sample size involved)
> of statistical significance - suggesting that on another week
> the result might have been 4-1 to Lee Sedol.
>
> I pointed out that in games of skill there's much more to judge than just
> the final
> outcome of each game, but wondered if anyone had any better (or worse :)
> arguments - or had even engaged in the same type of
> conversation.
>
> With AlphaGo winning 4 games to 1, from a simplistic
> stats point of view (with the prior assumption of a fair
> coin toss) you'd not be able to claim much statistical
> significance, yet most (me included) believe that
> AlphaGo is a genuinely better Go player than Lee Sedol.
>
> From a stats viewpoint you can use this approach:
> http://www.inference.phy.cam.ac.uk/itprnn/book.pdf
> (see section 3.2 on page 51)
>
> but given even priors it won't tell you much.
>
> Anyone know any good references for refuting this
> type of argument - the fact is of course that a game of Go
> is nothing like a coin toss.  Games of skill tend to base their
> outcomes on the result of many (in the case of Go many hundreds of)
> individual actions.
>
> Best wishes,
>
>   Simon
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] March KGS bot tournament - slow

[uurtamo .]

David,

It'd be cool if they were willing to spend a few $K to participate (Just a
guess about what the CPU time would cost). They've proven their strength (4
- 1 means easily pro strength, I think), so it would be a gift to the
community if they participated.

steve


On Thu, Mar 17, 2016 at 9:15 AM, David Doshay  wrote:

> Is there any way to forward this to the AlphaGo team? Comparing AlphaGo to
> the regular set of participants would be enlightening.
>
> Cheers,
> David G Doshay
>
> ddos...@mac.com
>
>
>
>
>
> On 17, Mar 2016, at 5:57 AM, Nick Wedd  wrote:
>
> The March KGS slow bot tournament will start on Sunday, March 27th, at 22:00
> UTC and end by 14:00 UTC on Wednesday 30th.  It will use 19x19 boards,
> with time limits of 235 minutes (almost four hours) each plus fast Canadian
> overtime, and komi of 7.5.  See http://www.gokgs.com/tournInfo.jsp?id=1020
>
> Please register by emailing me, with the words "KGS Tournament Registration"
> in the email title, at mapr...@gmail.com .
>
> Nick
> --
> Nick Wedd  mapr...@gmail.com
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] March KGS bot tournament - slow

[uurtamo .]

I like that: "the alphago"
On Mar 17, 2016 2:12 PM, "Lukas van de Wiel" 
wrote:

> Aja Huang is on the mailing list, and he is also on the compyer-go mailing
> list. So the AlphaGo is aware. :-)
>
> On Fri, Mar 18, 2016 at 9:29 AM, uurtamo .  wrote:
>
>> David,
>>
>> It'd be cool if they were willing to spend a few $K to participate (Just
>> a guess about what the CPU time would cost). They've proven their strength
>> (4 - 1 means easily pro strength, I think), so it would be a gift to the
>> community if they participated.
>>
>> steve
>>
>>
>> On Thu, Mar 17, 2016 at 9:15 AM, David Doshay  wrote:
>>
>>> Is there any way to forward this to the AlphaGo team? Comparing AlphaGo
>>> to the regular set of participants would be enlightening.
>>>
>>> Cheers,
>>> David G Doshay
>>>
>>> ddos...@mac.com
>>>
>>>
>>>
>>>
>>>
>>> On 17, Mar 2016, at 5:57 AM, Nick Wedd  wrote:
>>>
>>> The March KGS slow bot tournament will start on Sunday, March 27th, at
>>>  22:00 UTC and end by 14:00 UTC on Wednesday 30th.  It will use 19x19
>>> boards, with time limits of 235 minutes (almost four hours) each plus
>>> fast Canadian overtime, and komi of 7.5.  See
>>> http://www.gokgs.com/tournInfo.jsp?id=1020
>>>
>>> Please register by emailing me, with the words "KGS Tournament Registration"
>>> in the email title, at mapr...@gmail.com .
>>>
>>> Nick
>>> --
>>> Nick Wedd  mapr...@gmail.com
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>>
>>>
>>> ___
>>> Computer-go mailing list
>>> Computer-go@computer-go.org
>>> http://computer-go.org/mailman/listinfo/computer-go
>>>
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo & DCNN: Handling long-range dependency

[uurtamo .]

It's pretty incredible for sure.

s.
On Mar 14, 2016 2:20 PM, "Jim O'Flaherty" 
wrote:

> Whatever the case, a huge turn has been made and the next 5 years in Go
> are going to be surprising and absolutely fascinating. For a game that
> +2,500 years old, I'm beyond euphoric to be alive to get to witness this.
>
>
> On Mon, Mar 14, 2016 at 4:15 PM, Darren Cook  wrote:
>
>> > You can also look at the score differentials. If the game is perfect,
>> > then the game ends up on 7 points every time. If players made one
>> > small error (2 points), then the distribution would be much narrower
>> > than it is.
>>
>> I was with you up to this point, but players (computer and strong
>> humans) play to win, not to maximize the score. So a small error in the
>> opening or middle game can literally be worth anything by the time the
>> game ends.
>>
>> > I am certain that there is a vast gap between humans and perfect
>> > play. Maybe 24 points? Four stones??
>>
>> 24pts would be about two stones (if each handicap stone is twice komi,
>> e.g. see http://senseis.xmp.net/?topic=2464).
>>
>> The old saying is that a pro would need to take 3 to 4 stones against
>> god (i.e. perfect play).
>>
>> Darren
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo & DCNN: Handling long-range dependency

[uurtamo .]

Watching games 1-2 stones over you is helpful. There's some limit (9
stones? ) where it's hard to learn much,  but computers aren't (apparently)
there yet.

s.
On Mar 14, 2016 9:36 AM, "Jim O'Flaherty" 
wrote:

> I'm using the term "teacher" loosely. Any player who is better than me is
> an opportunity to learn. Being able to interact with the superior AI player
> strictly through actual play in a repeatable and undo-able form allows me
> to experiment and explore independently, in a way not achievable with a
> superior skilled human. This doesn't diminish the value of human teachers.
> In fact, I see them exploiting AIs to 'trivially play out all variations"
> when attempting to demonstrate why a particular move is desirable or
> undesirable.
>
> To support your point, though, I completely agree the kind of
> formalization you are pursuing will have even higher value as the AIs
> stretch out beyond humans. I think your work when combined with an untiring
> AI assistant will help human Go considerably. I know I have been helped
> greatly by your work, and I'm at a very amateur skill level.
>
> And your right about the more rigorous meaning of teacher (or Sensei)
> being quite a bit further away. I'm hopeful other AI breakthroughs outside
> of the Go domain will help close the gap more quickly.
> On Mar 14, 2016 9:21 AM, "Robert Jasiek"  wrote:
>
>> On 14.03.2016 08:59, Jim O'Flaherty wrote:
>>
>>> an AI player who becomes a better and better teacher.
>>>
>>
>> But you are aware that becoming a stronger AI player does not equal
>> becoming a stronger teacher? Teachers also need to (translate to and)
>> convey human knowledge and reasoning, and adapt to the specific pupils'
>> needs (incl. reasoning, subconscious thinking and psychology) while
>> interacting with human language specialised in go language. Solve two dozen
>> AI tasks, combine them and then, maybe, you get the equivalent of a
>> teacher. [FYI, I have taught 100+ regular single go pupils since 2008, and
>> groups of pupils.]
>>
>> --
>> robert jasiek
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo & DCNN: Handling long-range dependency

[uurtamo .]

There's a _whole_ lot of philosophizing going on on the basis of four
games.  Just saying.

steve
On Mar 14, 2016 7:41 AM, "Josef Moudrik"  wrote:

> Moreover, it might not be possible to explain the strong play in human
> understandable terms anyway; human rationalization might simply be a
> heuristic not strong enough to describe/capture it succinctly.
>
> On Mon, Mar 14, 2016 at 3:21 PM Robert Jasiek  wrote:
>
>> On 14.03.2016 08:59, Jim O'Flaherty wrote:
>> > an AI player who becomes a better and better teacher.
>>
>> But you are aware that becoming a stronger AI player does not equal
>> becoming a stronger teacher? Teachers also need to (translate to and)
>> convey human knowledge and reasoning, and adapt to the specific pupils'
>> needs (incl. reasoning, subconscious thinking and psychology) while
>> interacting with human language specialised in go language. Solve two
>> dozen AI tasks, combine them and then, maybe, you get the equivalent of
>> a teacher. [FYI, I have taught 100+ regular single go pupils since 2008,
>> and groups of pupils.]
>>
>> --
>> robert jasiek
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Finding Alphago's Weaknesses

[uurtamo .]

Not to put too fine a point on it, but there's not very many two or
three-move combos on an empty board. As staggering as it is, I'm inclined
to believe without further evidence that there's no book or just a very
light book.

s.
On Mar 10, 2016 7:50 PM, "Seo Sanghyeon"  wrote:

> 2016-03-11 11:42 GMT+09:00 terry mcintyre :
> > Hypothetically, they could have grafted one on. I read a report that the
> > first move in game 2 vs. Lee Sedol took only seconds. On the other hand,
> > it's first move in game 1 took a longer while. We can only speculate.
>
> This is easy to explain. AlphaGo was white (second to play) in game 1,
> and black (first to play) in game 2. You can precalculate a move if you are
> first to play. Harder to do that if you are second.
>
> --
> Seo Sanghyeon
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Finding Alphago's Weaknesses

[uurtamo .]

If that's the case, then they should be able to give opinions on best first
moves, best first two move combos, and best first three move combos. That'd
be interesting to see. (Top 10 or so of each).

s.
On Mar 10, 2016 12:37 PM, "Sorin Gherman"  wrote:

> From reading their article, AlphaGo makes no difference at all between
> start, middle and endgame.
> Just like any other position, the empty (or almost empty, or almost full)
> board is just another game position in which it chooses (one of) the most
> promising moves in order to maximize her chance of winning.
> On Mar 10, 2016 12:31 PM, "uurtamo ."  wrote:
>
>> Quick question - how, mechanically, is the opening being handled by alpha
>> go and other recent very strong programs? Giant hand-entered or
>> game-learned joseki books?
>>
>> Thanks,
>>
>> steve
>> On Mar 10, 2016 12:23 PM, "Thomas Wolf"  wrote:
>>
>>> My 2 cent:
>>>
>>> Recent strong computer programs never loose by a few points.  They are
>>> either
>>> crashed before the end game starts (because when being clearly behind
>>> they play more
>>> desperate and weaker moves because they mainly get negative feadback from
>>> their search with mostly loosing branches and risky play gives them the
>>> only
>>> winning sequences in their search) or they win by resignation or win
>>> by a few points.
>>>
>>> In other words, if a human player playing AlphaGo does not have a large
>>> advantage already in the middle game, then AlphaGo will win whether it
>>> looks
>>> like it or not (even to a 9p player like Michael Redmond was surprised
>>> last night about the sudden gain of a number of points by AlphaGo in the
>>> center in the end game: 4:42:10, 4:43:00, 4:43:28 in the video
>>> https://gogameguru.com/alphago-2/)
>>>
>>> In the middle and end game the reduced number of possible moves and the
>>> precise and fast counting ability of computer programs are superior.  In
>>> the
>>> game commentary of the 1st game it was mentioned that Lee Sedol
>>> considers the
>>> opening not to be his strongest part of the game.  But with AlphaGo
>>> playing
>>> top pro level even in the opening, a large advantage after the middle
>>> game
>>> might simply be impossible to reach for a human.
>>>
>>> About finding weakness:
>>> In the absense of games of AlphaGo to study it might be interesting to
>>> get a general idea by checking out the games where 7d Zen lost on KGS
>>> recently.
>>>
>>> Thomas
>>>
>>> On Thu, 10 Mar 2016, wing wrote:
>>>
>>> One question is whether Lee Sedol knows about these weaknesses.
>>>> Another question is whether he will exploit those weaknesses.
>>>> Lee has a very simple style of play that seems less ko-oriented
>>>> than other players, and this may play into the hands of Alpha.
>>>>
>>>> Michael Wing
>>>>
>>>>  I was surprised the Lee Sedol didn't take the game a bit further to
>>>>>  probe AlphaGo and see how it responded to [...complex kos, complex ko
>>>>>  fights, complex sekis, complex semeais, ..., multiple connection
>>>>>  problems, complex life and death problems] as ammunition for his next
>>>>>  game. I think he was so astonished at being put into a losing
>>>>>  position, he wasn't mentally prepared to put himself in a student's
>>>>>  role again, especially to an AI...which had clearly played much weaker
>>>>>  games just 6 months ago. I'm hopeful Lee Sedol's team has been some
>>>>>  meta-strategy sessions where, if he finds himself in a losing position
>>>>>  in game two, he turns it into exploring a set of experiments to tease
>>>>>  out some of the weaknesses to be better exploited in the remaining
>>>>>  games.
>>>>>
>>>>>  On Thu, Mar 10, 2016 at 8:16 AM, Robert Jasiek 
>>>>> wrote:
>>>>>
>>>>> >  On 10.03.2016 00:45, Hideki Kato wrote:
>>>>> > > >  such as solving complex semeai's and double-ko's, aren't solved
>>>>> yet.
>>>>> > >  To find out Alphago's weaknesses, there can be, in particular,
>>>>> > >  - this match
>>>>> >  - careful analysis of its games
>>>>> >  - Alphago playing on artificial 

Re: [Computer-go] Finding Alphago's Weaknesses

[uurtamo .]

Quick question - how, mechanically, is the opening being handled by alpha
go and other recent very strong programs? Giant hand-entered or
game-learned joseki books?

Thanks,

steve
On Mar 10, 2016 12:23 PM, "Thomas Wolf"  wrote:

> My 2 cent:
>
> Recent strong computer programs never loose by a few points.  They are
> either
> crashed before the end game starts (because when being clearly behind they
> play more
> desperate and weaker moves because they mainly get negative feadback from
> their search with mostly loosing branches and risky play gives them the
> only
> winning sequences in their search) or they win by resignation or win
> by a few points.
>
> In other words, if a human player playing AlphaGo does not have a large
> advantage already in the middle game, then AlphaGo will win whether it
> looks
> like it or not (even to a 9p player like Michael Redmond was surprised
> last night about the sudden gain of a number of points by AlphaGo in the
> center in the end game: 4:42:10, 4:43:00, 4:43:28 in the video
> https://gogameguru.com/alphago-2/)
>
> In the middle and end game the reduced number of possible moves and the
> precise and fast counting ability of computer programs are superior.  In
> the
> game commentary of the 1st game it was mentioned that Lee Sedol considers
> the
> opening not to be his strongest part of the game.  But with AlphaGo playing
> top pro level even in the opening, a large advantage after the middle game
> might simply be impossible to reach for a human.
>
> About finding weakness:
> In the absense of games of AlphaGo to study it might be interesting to get
> a general idea by checking out the games where 7d Zen lost on KGS
> recently.
>
> Thomas
>
> On Thu, 10 Mar 2016, wing wrote:
>
> One question is whether Lee Sedol knows about these weaknesses.
>> Another question is whether he will exploit those weaknesses.
>> Lee has a very simple style of play that seems less ko-oriented
>> than other players, and this may play into the hands of Alpha.
>>
>> Michael Wing
>>
>>  I was surprised the Lee Sedol didn't take the game a bit further to
>>>  probe AlphaGo and see how it responded to [...complex kos, complex ko
>>>  fights, complex sekis, complex semeais, ..., multiple connection
>>>  problems, complex life and death problems] as ammunition for his next
>>>  game. I think he was so astonished at being put into a losing
>>>  position, he wasn't mentally prepared to put himself in a student's
>>>  role again, especially to an AI...which had clearly played much weaker
>>>  games just 6 months ago. I'm hopeful Lee Sedol's team has been some
>>>  meta-strategy sessions where, if he finds himself in a losing position
>>>  in game two, he turns it into exploring a set of experiments to tease
>>>  out some of the weaknesses to be better exploited in the remaining
>>>  games.
>>>
>>>  On Thu, Mar 10, 2016 at 8:16 AM, Robert Jasiek  wrote:
>>>
>>> >  On 10.03.2016 00:45, Hideki Kato wrote:
>>> > > >  such as solving complex semeai's and double-ko's, aren't solved
>>> yet.
>>> > >  To find out Alphago's weaknesses, there can be, in particular,
>>> > >  - this match
>>> >  - careful analysis of its games
>>> >  - Alphago playing on artificial problem positions incl. complex kos,
>>> >  complex ko fights, complex sekis, complex semeais, complex endgames, >
>>> multiple connection problems, complex life and death problems (such as >
>>> Igo Hatsu Yoron 120) etc., and then theoretical analysis of such play
>>> >  - semantic verification of the program code and interface
>>> >  - theoretical study of the used theory and the generated dynamic data
>>> >  (structures)
>>> > >  --
>>> >  robert jasiek
>>> >  ___
>>> >  Computer-go mailing list
>>> >  Computer-go@computer-go.org
>>> >  http://computer-go.org/mailman/listinfo/computer-go [1]
>>>
>>>
>>>
>>>  Links:
>>>  --
>>>  [1] http://computer-go.org/mailman/listinfo/computer-go
>>>
>>>  ___
>>>  Computer-go mailing list
>>>  Computer-go@computer-go.org
>>>  http://computer-go.org/mailman/listinfo/computer-go
>>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Other German poll on Lee Sedol vs AlphaGo

[uurtamo .]

Well, certainly they'd ship it to a kind German proxy. Friend of the court,
so to speak.

:)

s.
On Feb 29, 2016 11:10 PM, Ingo Althöfer <3-hirn-ver...@gmx.de> wrote:

> Hello,
>
> > For those who want to try their luck at the nice board, even without
> > knowing German:
> >
> > http://www.go-baduk-weiqi.de/gewinnspiel-lee-sedol-gegen-alphago/
>
> my impression is that Hebsacker Verlag will not be shipping
> the prize to outside Germany.
>
> But your (international) voting may be interesting nevertheless.
>
> Ingo.
>
>
> > -- Gonçalo
> >
> > On 01/03/2016 01:29, "Ingo Althöfer" wrote:
> > > Hebsacker-Verlag is the leading German provider of Go equipment.
> > > They have initiated a poll on the forthcoming man-machine match.
> > >
> > > So far 278 people have voted. The distribution of expected
> > > scores is:
> > >
> > > Lee Sedol wins by 5-0: 34.2 %
> > > Lee Sedol wins by 4-1: 29.1 %
> > > Lee Sedol wins by 3-2: 12.2 %
> > > total: 75.5 %
> > >
> > > AlphaGo wins by 3-2: 7.2 %
> > > AlphaGo wins by 4-1: 9.4 %
> > > AlphaGo wins by 5-0: 7.9 %
> > > total: 24.5 %
> > >
> > > Amongst those "predicting the correct score" a nice
> > > go board (valued 380 Euro) is raffled.
> > >
> > > Ingo.
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] AlphaGo and the Standard Mistake in Research and Journalism

[uurtamo .]

Not to beat a dead horse, but big numbers aren't inherently interesting to
describe.

There are integers bigger than any integer anyone has written down in any
form. This particular integer is large, but "consumable".

I guess I get tired of the "number of atoms in the observable universe"
comparison. Plenty of integers this size or larger can be dealt with in
other ways. (Think factoring).

steve


On Mon, Feb 1, 2016 at 2:11 AM, Olivier Teytaud 
wrote:

>
>> How do you know that an implicit expression (of length smaller than
>> 10^80) of the number does not exist? :)
>>
>
> I am pretty sure that such an implicit expression exists: it is << the
> number of etc etc >> (formalized for your favorite set of rules :-) ).
>
>
>
>
> --
> =
> Olivier Teytaud, INRIA TAO Research Fellow ---
> http://www.slideshare.net/teytaud
> "Please stop quoting me on internet."___ Albert Einstein
>
>
>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Mathematics in the world

[uurtamo .]

Robert,

Just as an aside, I really respect your attention to detail and your
insistence that proof technique follow generalizing statements about
aspects of go.

I think that the counting problems recently were pretty interesting (number
of positions versus number of games).

The engineering problem of winning against humans is of course much simpler
than the math problem of understanding the game deeply (just think about
all of the work Berlekamp did working on this). I like that they move
together hand-in-hand, and right now it seems like the human urge for most
people is to make a set of computers strong enough that they can beat a top
pro, once, then the next goal will be to beat them regularly, as has been
done in chess.

If you think about chess endgames, and how they've been categorized, we're
nowhere near that in Go except for fairly quiescent positions. The opening
is a nightmare, the midgame is a nightmare, and multiple fights with
multiple kos are a nightmare. Solving this mathematically is of course
hugely far in the future. Faking your way forward with engineering (similar
to how people play) seems to be our best guess at the moment.

Thanks for your insight and rigor and I'm glad that you're continuing down
your rigorous path when so many of us have forgotten that extremely minor
rule differences can be: inexplicable (japanese rules, if i understand
correctly), difficult to deal with (chinese rules, under very liberal
understanding) or useless (mathematical descriptions of a game which is
totally different than how people actually play).

Thanks again,

steve


On Tue, Feb 2, 2016 at 10:54 AM, Robert Jasiek  wrote:

> On 02.02.2016 13:05, "Ingo Althöfer" wrote:
>
>> when a student starts
>> studying Mathematics (s)he learns in the first two semesters that
>> everything has to be defined waterproof. Later, in particular
>> when (s)he comes near to doing own research, you have to make
>> compromises - otherwise you will never make much progress.
>>
>
> When I studied maths and theoretical informatics at FU Berlin (and a bit
> at TU Berlin) (until quitting because of studying too much go, of course),
> during all semesters with every paper, lecture, homework or professor,
> everything had to be well-defined, assumptions complete and mandatory
> proofs accurate.
>
> As a hobby go theory / go rules theory researcher, I can afford the luxury
> of choosing formality (see Cycle Law), semi-formality (see Ko) or
> informality (in informal texts) because I need not pass university degrees
> with the work. My luxury of laziness / convenience when I use semi-formal
> style (as typical in the theory parts of my go theory papers) indeed has
> the advantages of being understood more easily from the go player's (also
> my own) perspective and allowing my faster research progress. If I had had
> to use formal style for every text, I might have finished only half of the
> papers.
>
> If we can believe Penrose (The Road to Reality) and Smolin (The Trouble
> with Physics), the world of mathematical physics is split into guesswork
> (string theory without valid mathematical foundation) and accurate maths.
> Progress might not be made because too many have lost themselves in the
> black hole of ambiguous string theory. Computer go theory seems to be
> similar to physics.
>
> --
> robert jasiek
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] What hardware to use to train the DNN

[uurtamo .]

David,

You're a trooper for doing this in windows. :)

The OS overhead is generally lighter if you use unix; even the most modern
windows versions have a few layers of slowdown. Unix (for better or worse)
will give you closer, easier access to the hardware, and closer, easier
access to halting your machine if you are deep in the guts. ;)

s.


On Tue, Feb 2, 2016 at 10:25 AM, David Fotland 
wrote:

> Detlef, Hiroshi, Hideki, and others,
>
> I have caffelib integrated with Many Faces so I can evaluate a DNN.  Thank
> you very much Detlef for sample code to set up the input layer.  Building
> caffe on windows is painful.  If anyone else is doing it and gets stuck I
> might be able to help.
>
> What hardware are you using to train networks?  I don’t have a
> cuda-capable GPU yet, so I'm going to buy a new box.  I'd like some
> advice.  Caffe is not well supported on Windows, so I plan to use a Linux
> box for training, but continue to use Windows for testing and development.
> For competitions I could use either windows or linux.
>
> Thanks in advance,
>
> David
>
> > -Original Message-
> > From: Computer-go [mailto:computer-go-boun...@computer-go.org] On Behalf
> > Of Hiroshi Yamashita
> > Sent: Monday, February 01, 2016 11:26 PM
> > To: computer-go@computer-go.org
> > Subject: *SPAM* Re: [Computer-go] DCNN can solve semeai?
> >
> > Hi Detlef,
> >
> > My study heavily depends on your information. Especially Oakfoam code,
> > lenet.prototxt and generate_sample_data_leveldb.py was helpful. Thanks!
> >
> > > Quite interesting that you do not reach the prediction rate 57% from
> > > the facebook paper by far too! I have the same experience with the
> >
> > I'm trying 12 layers 256 filters, but it is around 49.8%.
> > I think 57% is maybe from KGS games.
> >
> > > Did you strip the games before 1800AD, as mentioned in the FB paper? I
> > > did not do it and was thinking my training is not ok, but as you have
> > > the same result probably this is the only difference?!
> >
> > I also did not use before 1800AD. And don't use hadicap games.
> > Training positions are 15693570 from 76000 games.
> > Test positions are   445693 from  2156 games.
> > All games are shuffled in advance. Each position is randomly rotated.
> > And memorizing 24000 positions, then shuffle and store to LebelDB.
> > At first I did not shuffle games. Then accuracy is down each 61000
> > iteration (one epoch, 256 mini-batch).
> > http://www.yss-aya.com/20160108.png
> > It means DCNN understands easily the difference 1800AD games and  2015AD
> > games. I was surprised DCNN's ability. And maybe 1800AD games  are also
> > not good for training?
> >
> > Regards,
> > Hiroshi Yamashita
> >
> > - Original Message -
> > From: "Detlef Schmicker" 
> > To: 
> > Sent: Tuesday, February 02, 2016 3:15 PM
> > Subject: Re: [Computer-go] DCNN can solve semeai?
> >
> > > Thanks a lot for sharing this.
> > >
> > > Quite interesting that you do not reach the prediction rate 57% from
> > > the facebook paper by far too! I have the same experience with the
> > > GoGoD database. My numbers are nearly the same as yours 49% :) my net
> > > is quite simelar, but I use 7,5,5,3,3, with 12 layers in total.
> > >
> > > Did you strip the games before 1800AD, as mentioned in the FB paper? I
> > > did not do it and was thinking my training is not ok, but as you have
> > > the same result probably this is the only difference?!
> > >
> > > Best regards,
> > >
> > > Detlef
> >
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Knowledge Details

[uurtamo .]

Just as an aside,

One nice thing about having "expert" chess players is the ability to easily
discover cheating and to estimate the "player rank" of any move. Because
the computer is effectively an oracle for that game, it gives incidental
feedback about strength of any given move.

steve
On Feb 3, 2016 6:34 AM, "Jim O'Flaherty"  wrote:

> Robert,
>
> How have these things emerged in the chess AI world following Deep Blue
> and Kasperov's loss over a decade ago? To what degree does "human expert
> details of chess theory matters" (where the term "matters" is pretty
> squishy). From what I can see, that is not what happened and while I am not
> privy to every detail of every motivation in the chess AI world, I'm
> certainly not seeing this assertion or the supporting values arising to any
> level of relevance, much less primacy.
>
> And for those who were working in the Chess knowledge world, how was their
> work, business, grants and funding affected by the Deep Blue/Kasperov
> results and then the rapid improvement of more than one chess engine to
> beyond the highest skilled humans? What happened to prize tournaments? To
> what degree is it reasonable to predict a similar pattern will occur in and
> about Go and those who are working in the Go knowledge world?
>
> BTW, I have my own personal aspirations which have been thwarted by this
> development. I have several thousand hours of doing my own research and
> development (of my personal spare time outside my day job, over many years)
> which has been rendered considerably less valuable (other than my own
> personal development in the non-Go related parts). And I'm finding it
> difficult to embrace this "change" as I had no idea just how much
> motivation it created in the present having the Go AI goal as an inspiring
> future. The loss of that motivation has created anxiety and uncertainty.
>
> And even in spite of the loss and the grief I am experiencing in that
> loss, I am still very enthusiastic about Aja and his team's achievements.
> And I will be following all the teams who continue to work in this area.
> For myself, I will now look for other ways to apply my knowledge, although
> I will likely drift further away from Go as the focal point of motivation.
>
> Best of luck finding your way through your meaning and value (emotional)
> reintegration of this newest reality update.
>
>
> Namaste,
>
> Jim
>
>
> On Wed, Feb 3, 2016 at 3:51 AM, Robert Jasiek  wrote:
>
>> The current fashion favours general AI approaches forgoing knowledge
>> details. Given enough calculation power applied to well chosen AI
>> techiques, many knowledge details are redundant because they are generated
>> automatically: AlphaGo does play (at least some) ko fights with ko threats,
>> tesujis, test moves, (at least some) life and death or semeai problems etc.
>> At the same time, AI calculation power is still not large enough to
>> generate all human knowledge details. Aji with long-term impact and
>> maintaining the life status "independently alive" instead of unnecessarily
>> transforming it to "(ko|independently alive)" (aka "unsettled") are prime
>> examples. Programs also play for the win regardless of whether moves are
>> suboptimal for the score difference - human players tend to avoid such
>> (programs would also profit from avoiding such to prevent losing when
>> making a later mistake due to a knowledge gap related to insufficient error
>> handling). There is another great threat related to knowledge details,
>> which is not immediately apparent and will be even much less apparent when
>> programs will exceed top human playing strength: A program can run into a
>> situation where an infrequent knowledge detail becomes relevant. And a
>> program can run into ordinary software or hardware bugs, something that
>> must be detected and correct on the AI level.
>>
>> My conclusion is: human expert knowledge on details of go theory matters.
>>
>> There have been 9p players committing self-atari when filling a dame, so
>> you might argue that programs may infrequently make similar blunders. When
>> I issued a million dollar prize, I'd prefer human expert knowledge
>> implemented at least as an additional layer of error handling.
>>
>> (Other fun includes internet connection trouble, server bugs of
>> distributed computers, hardware bugs of the local interface computers or
>> interrupted power supply.)
>>
>> --
>> robert jasiek
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Computer-go Digest, Vol 72, Issue 41

[uurtamo .]

It might even be interesting if it took place *before* the lee sedol match.

s.
On Jan 31, 2016 5:09 PM, "Chaohao Pan"  wrote:

> Just in case that no one knows it. Ke Jie has publicly announced that he
> is willing to play against AlphaGo, even without any prize money. Since Ke
> Jie is absolutely the current No.1, it would be a good choice to have
> another match with Ke Jie, time permitting, no matter AlphaGo wins or loses
> against Lee Sedol,.
>
> Chaohao
>
> 2016-01-31 13:34 GMT-08:00 John Tromp :
>
>> > You must be kidding about Lee Sedol.
>> > ...
>> > So he was by far the biggest fish Google could ever catch for that
>> > game, for Go insiders as well as for people outside the Go scene.
>>
>> Well said, Marc.
>>
>> In terms of name recognition and domination in the past decade,
>> who else but Lee Sedol should be picked as the "Kasparov of Go"
>> in the ultimate Man vs Machine match?
>>
>> -John
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Those were the days ...

[uurtamo .]

That might be a stone or two difference for that software at best, no?

s.
On Dec 28, 2015 6:58 AM, "Michael Sué"  wrote:

> > Remember 1998: In the US Go Congress an exhibition match
> > took place: 5-dan Martin Mueller against Many Faces of Go.
> > Martin gave 29 handicap stones - and won "handily".
> > 29 stones - can you believe it?
>
> Martin Mueller, 5d, won an H-29 game against a 1998 go-program that ran
> run on a 1998 computer. So, in order to reproduce the playing strength of
> the old program and get a fair comparison it should again run on an old
> machine while the modern go-programs use today's hardware.
> - Michael.
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] NHK News "Bots will be admitted to pro tournaments"

[uurtamo .]

This would be quite amazing.

s.
On Dec 23, 2015 5:22 AM, "甲斐徳本"  wrote:

> NHK (BBC equivalent in Japan) reported in tonight's 7 o'clock news on
> national TV channel that "Nihon Kiin will be admitting computer programs to
> pro tournaments as participants"
>
> *Details are not known at all.  It is a common practice for Nihon Kiin to
> let its sponsors (newspapers, TV / cable channels, etc) scoop valuable news
> before making formal announcements.  For example, Antti* Törmänen's 1p
> status was reported by Mainichi newspaper a day before Nihon Kiin's
> announcement on its website.
>
> *So let's hope some detail will become available tomorrow.*
>
> *Tokumoto*
>
>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Maximum Frequency method

[uurtamo .]

BTW: have you tried other distributional difference metrics, or does K-L
have properties that you like?

Thanks,

steve
On Sep 5, 2015 1:35 AM, "Hideki Kato"  wrote:

> djhbrown .: <
> capsify9fub60pd3lzdyhdpupffgyenv4t+m47okwphzrb4q...@mail.gmail.com>:
> >thank you for sharing the paper.
> >
> >"the Maximum Frequency method is based on the
> >maximization of the difference between the expected reward of
> >the optimal move and that of others"
> >
> >intuitively it feels that biasing random search towards the optimal route
> >would yield reduced failure rates, yet it does seem to depend on knowing
> >what the optimal route is beforehand.
>
> UCT is never a random search but deterministic.
>
> Maxmizing KL-divergence just speed-up the convergence of the interative
> algorithm.
>
> Hideki
>
> >if i knew the optimal route to get from A to B, i wouldn't bother doing a
> >random search, but just follow it.
> >
> >"This property [“bias in suboptimal moves”] means that the impact of
> >missing the optimal move is much greater for one player than it is for the
> >opponent."
> >
> >i find this conclusion puzzling because Go is a zero-sum game, so what is
> >good for one side is equally bad for the other, not variably so.  I have
> >not checked the statistical inference calculations to see whether there is
> >an error in them.
> > inline file
> >___
>
> >Computer-go mailing list
>
> >Computer-go@computer-go.org
>
> >http://computer-go.org/mailman/listinfo/computer-go
> --
> Hideki Kato 
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] re comments on Life and Death

[uurtamo .]

Learned rules from pure stats might be good guiding posts, but the pure
checking of millions of board positions is always going to be necessary.

My $0.02,

s.
On Sep 4, 2015 3:49 PM, "Jim O'Flaherty"  wrote:

> I disagree with the assertion MC must be the starting point. It appears to
> have stagnated into a local optima; i.e. it's going to take something
> different to dislodge MC, just like it took MC to dislodge the traditional
> approaches preceding MC's introduction a decade ago. Ultimately, I think it
> can serve to inform a higher level conceptual system
>
> And while I don't get his videos (they are way to ADHD scattered and
> discontinuous for my personal ability to focus and internalize), I think I
> grok the general direction he'd like to see things head. And I am quite
> ambivalent about the idea of creating and using linguistic semantic trees
> as an approach, as much or even more than I was about MC when it emerged.
>
> On Fri, Sep 4, 2015 at 10:55 AM, Stefan Kaitschick <
> stefan.kaitsch...@hamburg.de> wrote:
>
>> So far I have not criticised but asked questions. I am a great fan of the
>> expert system approach because a) I have studied go knowledge a lot and
>> see, in principle, light at the end of the tunnel, b) I think that "MC +
>> expert system" or "only expert system" can be better than MC if the expert
>> system is well designed, c) an expert system can, in principle, provide
>> more meaningful insight for us human duffers than an MC because the expert
>> system can express itself in terms of reasoning. (Disclaimer: There is a
>> good chance that I will criticise anybody presenting his definitions for
>> use in an expert system. But who does not dare to be criticised does not
>> learn!)
>>
>> MC is currently stagnating, so looking at new (or old discarded)
>> approaches has become more attractive again.
>> But I don't think that a "classic" rules based system will be of much use
>> from here. It is just too far removed from MC concepts to be productively
>> integrated into an MC system. And no matter what, MC has to be the starting
>> point, because it is so much more effective than anything else that has
>> been tried.What you are left to work with, is the trail of statistics that
>> MC leaves behind. That is the only tunnel with a possible end to it that I
>> see. And who knows, maybe someone will find statistical properties that can
>> be usefully mapped back to human concepts of go.
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] EGC2015 Events

[uurtamo .]

Petr,

Thanks. Is Haylee's game on kgs, or only youtube?

Thanks in advance,

steve


On Wed, Jul 29, 2015 at 11:21 AM, Petr Baudis  wrote:

>   Hi!
>
>   There are several Computer Go events on EGC2015.  There was a small
> tournament of programs, played out on identical hardware by each,
> won by Aya:
>
> https://www.gokgs.com/tournEntrants.jsp?sort=s&id=981
>
>   Then, one of the games, Aya vs. Many Faces, was reviewed by Lukas
> Podpera 6d:
>
> https://www.youtube.com/watch?v=_3Lk1qVoiYM
>
>   Right now, Hajin Lee 3p (known for her live commentaries on Youtube
> as "Haylee") is playing Aya (giving 5 stones) and commenting live:
>
> https://www.youtube.com/watch?v=Ka2ilmu7Eo4
>
> --
> Petr Baudis
> If you have good ideas, good data and fast computers,
> you can do almost anything. -- Geoffrey Hinton
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Hex is solved ?

[uurtamo .]

Thanks.

Generally is good if it's in a reputable journal or conference. Not
everyone can do that, but if it's a new and interesting result of interest
to more than 100 people, probably you can. There are peer-reviewed
electronic journals, for instance.

It's also unclear what the formal statement of the claim is. Having an
efficient solution from an arbitrary board position for an arbitrary
boardsize isn't credible, as John pointed out. (And a version of that
result, which was published in acta informatica, isn't mentioned or
referenced on the hex page mentioned as far as I can find - it would be a
much bigger result than an *inefficient* algorithm, which is fairly easy.).

s.
I allowed my self to remove the link from the wikipedia page as non pair
reviewed

2015-07-29 15:53 GMT+02:00 "Ingo Althöfer" <3-hirn-ver...@gmx.de>:

> Hi Rèmi,
>
> gorget it - no serious work.
>
> Ingo.
>
>
> > Gesendet: Dienstag, 28. Juli 2015 um 15:38 Uhr
> > Von: "Rémi Coulom" 
> > An: computer-go@computer-go.org
> > Betreff: [Computer-go] Hex is solved ?
> >
> > Hi,
> >
> > I have just been told by a colleague that Edouard Rodrigues solved hex
> mathematically. I was very surprised because I had never heard about it.
> >
> > The web site with the proof and optimal strategy is there:
> > http://jeudhex.com/?page_id=17
> >
> > I did not look at it in details, but it seems his method can find an
> optimal move on any position and any board size.
> >
> > Did the computer-hex people of this list knew about it? I know there was
> an Hex tournament in Leiden, so I suppose the computer Hex community might
> not be aware of this result. Or maybe the mathematical result is wrong?
> >
> > Please circulate this information to the Hex specialists. I am curious
> about their opinion.
> >
> > I'll take time tomorrow to study that web site a little.
> >
> > Rémi
> > ___
> > Computer-go mailing list
> > Computer-go@computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>


___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Hex is solved ?

[uurtamo .]

Being pspace complete would just make the method impracticably slow for a
sufficiently large board size.

The searches in such a case will be exponential in board size and not be
very interesting.

s.
On Jul 28, 2015 6:59 AM, "John Tromp"  wrote:

> > I have just been told by a colleague that Edouard Rodrigues solved hex
> mathematically. I was very surprised because I had never heard about it.
> >
> > The web site with the proof and optimal strategy is there:
> > http://jeudhex.com/?page_id=17
>
> Perhaps he found a winning strategy for an unrestricted first player?
> The game with the usual swap rule doesn't feel to me like it would submit
> to
> an efficiently computable strategy.
>
> > I did not look at it in details, but it seems his method can find an
> optimal move on any position and any board size.
>
> That's most unlikely, considering that HEX is PSPACE complete...
>
> -John
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] CGT in Clobber ?!

[uurtamo .]

Well...

Not to put too fine a point on it, but if you show up to a vespa rally with
a hog, you'll win all long-distance races.

Regardless, there can be a lot of merit in the design decisions made in a
vespa.

(Even ones that might work on a Harley. :)

steve
On Jul 14, 2015 12:45 AM, Ingo Althöfer <3-hirn-ver...@gmx.de> wrote:

> Hello Josef, hello all,
>
> "Josef Moudrik" 
> > ... As far as I know, combinatorial game theory is not used
> > in modern Go engines, despite its nice theoretical properties.
>
> Let me tell you an anecdote from CGT history: In Februar 2002, there
> was a week-long conference on "Algorithmic Combinatorial Game Theory"
> in "Schloss Dagstuhl" (Germany). It was years before the Monte Carlo
> revolution, and in those days many hopes for strong go bots were
> on CGT.
>
> In the beginning of the week, the simple board game Clobber
> (perfectly suited for use of CGT) was introduced - and a human
> Clobber tournament was planned for the evenings.
> I proposed to write a Clobber bot (without using CGT), and
> two other participants followed me. Within hours, our bots
> were ready. And they crushed the humans (on 6x5 board size)
> terribly. So, "we" were excluded from the evening tournament
> (and had to play our own bot competition).  The only comment on
> my bot by senior Elwyn Berlekamp (one of the fathers of modern
> CGT; co-author of "Winning ways") was: Why does it have such an
> ugly graphical interface?  It seemed, Prof. Berlekamp was dis-
> appointed by the fact that Clobber bots were so strong without
> using any CGT.
>
> Photo with (almost) the whole group of 2002-participants:
> https://www.dagstuhl.de/Gruppenbilder/02081.A.B.JPG
> In the very first row (sitting): third person from left
> is John Tromp (the man from the Tromp-Taylor rules, from
> the Tromp-Cook bet, from go position counting and and and).
> In the same row the man with the diagonal go board is
> Martin Mueller. Directly behind Martin stands Richard Nowakowski
> (with red shirt), co-inventor of Clobber (in Summ,er 2001, together
> with M.H. Albert and J.P. Grossman).
>
> https://www.dagstuhl.de/en/program/calendar/semhp/?semnr=02081
>
> Looking forward to meet you in Liberec!
> Ingo.
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Strange GNU Go ladder behavior

[uurtamo .]

So what is the 64-bit problem? (Or did I misread?)
On Jun 19, 2015 8:04 PM, "Peter Drake"  wrote:

> Okay, that worked (with the correction that "ibstdc" should be "libstdc").
>
> The new version doesn't choke on my sgf file!
>
> Now for the acid test, running the whole experiment...
>
> On Fri, Jun 19, 2015 at 4:43 PM, Hiroshi Yamashita 
> wrote:
>
>> Configure doesn't seem happy with that:
>>>
>>
>> I'm not familiar with CentOS, but maybe need to install
>>
>> # yum -y install glibc-devel.i386 libstdc++-devel.i386
>> or
>> # yum -y install glibc-devel.i686 glibc-devel ibstdc++-devel.i686
>>
>> Hiroshi Yamashita
>>
>>
>> ___
>> Computer-go mailing list
>> Computer-go@computer-go.org
>> http://computer-go.org/mailman/listinfo/computer-go
>>
>
>
>
> --
> Peter Drake
> https://sites.google.com/a/lclark.edu/drake/
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Strange GNU Go ladder behavior

[uurtamo .]

What does it do for memory management? Is it ungracefully failing while
evaluating the ladder itself due to ram issues?

steve
On Jun 18, 2015 12:15 PM, "Peter Drake"  wrote:

> This list may not be able to help, but I'm running out of clues on this
> one.
>
> I'm trying to run an experiment playing Orego against GNU Go in many
> games. Some of the games don't end properly. As far as I can tell, here's
> what happens:
>
> 1) Orego plays out a losing ladder. (That needs to be fixed, but that's
> another problem.)
> 2) At some point in the ladder, GNU Go quietly dies without responding to
> the move request.
>
> Has anyone else encountered this?
>
> --
> Peter Drake
> https://sites.google.com/a/lclark.edu/drake/
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] evaluating number of wins versus looses

[uurtamo .]

For Wilson, you can use depth to pick confidence bound.

s.
On Mar 30, 2015 7:09 AM, "Petr Baudis"  wrote:

> On Mon, Mar 30, 2015 at 09:11:52AM -0400, Jason House wrote:
> > The complex formula at the end is for a lower confidence bound of a
> > Bernoulli distribution with independent trials (AKA biased coin flip) and
> > no prior knowledge. At a leaf of your search tree, that is the most
> correct
> > distribution. Higher up in a search tree, I'm not so sure that's the
> > correct distribution. For a sufficiently high number of samples, most
> > averaging processes converge to a Normal distribution (due to central
> limit
> > theorem). For a Bernoulli distribution with a mean near 50% the required
> > number of samples is ridiculously low.
> >
> > I believe a lower confidence bound is probably best for final move
> > selection, but UCT uses an upper confidence bound for tree exploration. I
> > recommend reading the paper, but it uses a gradually increasing
> confidence
> > interval which was shown to be an optimal solution for the muli-armed
> > bandit problem. I don't think that's the best model for computer go, but
> > the success of the method cannot be denied.
> >
> > The strongest programs have good "prior knowledge" to initialize wins and
> > losses. My understanding is that they use average win rate directly
> > (incorrect solution #2) instead of any kind of confidence bound.
> >
> > TL;DR: Use UCT until your program natures
>
> The strongest programs often use RAVE or LGRF or something like that,
> with or without the UCB for tree exploration.
>
> For selecting the final move, the move with most simulations is used.
> (Using the product reviews analogy - assume all your products go on sale
> at once, have the same price, shipping etc., then with number of buyers
> going to infinity, the best product should get the most buyers and
> ratings even if some explore other products.)  I think trying the Wilson
> lower bound could be also interesting, but the inconvenience is that you
> need to specify some arbitrary confidence level.
>
> > On Mar 30, 2015 8:06 AM, "folkert"  wrote:
> > > --
> > > Finally want to win in the MegaMillions lottery? www.smartwinning.info
>
> funny in the context :)
>
> --
> Petr Baudis
> If you do not work on an important problem, it's unlikely
> you'll do important work.  -- R. Hamming
> http://www.cs.virginia.edu/~robins/YouAndYourResearch.html
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] What's a good playout speed?

[uurtamo .]

I can offer you a factor of 2 speedup...

s.
On Mar 28, 2015 7:59 PM, "hughperkins2"  wrote:

> By the way, for mcts you dont need time controls. Each move takes the same
> amount of time, since you just do n playouts, and choose n as you like.
>
> I think my playouts took 2s, which was enoufht for 4 playouts i
> suppose, but it was in novemebr, dont remember clearly... It was obvious
> that adding more playouts didnt increase the strength libearly, more like
> logarithmically, or, at best as the square root, and i dont want to program
> heursitics by hand, not going to get a papet out of that :-)
>
> The atari paper looks interesting. Kind of dabbling in that a bit... By
> the way, for mcts you dont need time controls. Each move takes the same
> amount of time, since you just do n playouts, and choose n as you like.
>
> I think my playouts took 2s, which was enoufht for 4 playouts i
> suppose, but it was in novemebr, dont remember clearly... It was obvious
> that adding more playouts didnt increase the strength libearly, more like
> logarithmically, or, at best as the square root, and i dont want to program
> heursitics by hand, not going to get a papet out of that :- out of that :-)
>
> The atari paper looks interesting. Kind of dabbling in that a bit...
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] What's a good playout speed?

[uurtamo .]

Can you put in the game description: "really bad, only play if you're
patient" and put in quicker time controls?

s.
On Mar 28, 2015 3:25 PM, "hughperkins2"  wrote:

> You can name name a specific opponent, and then your bot will play against
> it.
>
> Automatch works, but tends to result in lots of people being forced to
> play your bot, and then leaving the game, after the bot took ages to play
> in some ridiculous location, which is kind of embarrassing :-P
>
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] implementing simple ko

[uurtamo .]

You can have multiple kos on the board. Two is the most usual case that
doesn't trigger other rules. Easiest (logically, not practically, perhaps)
is to never repeat a board position ever (so called superko, I think),
which could be implemented as a full-board-position hash.

s.
On Mar 11, 2015 10:08 PM, "Ray Tayek"  wrote:

> i need to implement a simple ko rule.
>
> looks like you need to keep track of 2 points for each ko and who took it
> last. so it looks like a list of (point,point,who) that can change each
> turn?
>
> any pointers will be appreciated.
>
> thanks
>
> --
> Honesty is a very expensive gift. So, don't expect it from cheap people -
> Warren Buffet
> http://tayek.com/
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go