Re: [Computer-go] mini-max with Policy and Value network

[Ingo Althöfer]

Hi, just my 2 Cent.

"Gian-Carlo Pascutto"  wrote:

 
> In the attached SGF, AlphaGo played P10, which was considered a very
> surprising move by all commentators...
> I can sort-of confirm this:
> 
> 0.295057654 (E13)
> ...(60 more moves follow)...
> 0.11952 (P10)
> 
> So, 0.001% probability. Demis commented that Lee Sedol's winning move in
> game 4 was a one in 10 000 move. This is a 1 in 100 000 move.

In Summer 2016 I checked the games of AlphaGo vs Lee Sedol
with repeated runs of CrazyStone DL:
In 3 of 20 runs the program selected P10. It
turned out that a rather early "switch" in the search was
necessary to arrive at P10. But if CS did that it
remained with this candidate.

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

Re: [Computer-go] What was the final score after the counting of AlphaGo-vs-Ke Jie Game #1?

[Hideki Kato]

No.  With Chinese rules and 7.5 komi, the result is W+0.5.  With 
Japanese rules and 7.5 komi, the result is W+1.5. Because black 
played the final move in the game (so the number of black stones 
is one more than white).

Hideki

Michael Alford: :
>Half point is the result using Japanese komi 6.5, Chinese komi is 7.5, 
>end result is W +1.5
>
>
>On 5/23/17 12:02 AM, Jim O'Flaherty wrote:
>> I have now heard that AlphaGo one by 0.5 points.
>>
>>
>> On Tue, May 23, 2017 at 2:00 AM, Jim O'Flaherty 
>> > wrote:
>>
>> The announcer didn't have her mic on, so I couldn't hear the final
>> score announced...
>>
>> So, what was the final score after the counting of AlphaGo-vs-Ke
>> Jie Game #1?
>>
>>
>>
>>
>> ___
>> 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

Re: [Computer-go] mini-max with Policy and Value network

[Hideki Kato]

Alvaro Begue: 

[Computer-go] Alphago

[Detlef Schmicker]

Hi,

might be a little impolite, but I wonder about the strength of alphago.

The version playing Ke Jie seems to be about as strong (or stronger) as
Ke Jie is. I have the feeling the playing strength is carefully chosen
not to be too strong.

In the press conference it was told, alphago is running on one machine
this time with 10% of the computational power of the version playing Lee
Sedol.

The thinking time seems to be quite low too, at least twice the time per
move should be no problem.

I wonder, if alphago still scales good with computational power. The
(not so) strong go programs on CGOS seem all still scale quite good with
computational power, about 100ELO stronger with twice of the
computational power.

Probably alphago team tested this and they know, if a version with 20
times more computational power can give the actual version really 3
stones, and so might be able to give 3 stones to very strong pros too?!


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

[Computer-go] Functioning?

[Michael Alford]

I haven't seen a post to the group in three days.  That can't be 
possible, with the event in China.

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

[Computer-go] winning probability on empty board

[Harald Bögeholz]

Hi,


I heard at the press conference after the second match that Ke Jie has
asked to take white for today's match against AlphaGo because he thinks
his chances of winning are better.

I am wondering what AlphaGo's evaluation of the empty board would be.
Does anybody know? How about other strong programs?


Harald

-- 
Harald Bögeholz (PGP key available from servers)

   int f[9814],b,c=9814,g,i;long a=1e4,d,e,h;
   main(){for(;b=c,c-=14;i=printf("%04d",e+d/a),e=d%a)
   while(g=--b*2)d=h*b+a*(i?f[b]:a/5),h=d/--g,f[b]=d%g;}
  (Arndt/Haenel)

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

[Computer-go] AlphaGo won second game against Ke Jie, 2-0

[Hiroshi Yamashita]

(;GM[1]SZ[19]
PB[AlphaGo]
PW[Ke Jie]
DT[2017-05-25]
RE[B+R]
KM[7.5]TM[180]RU[Chinese]PC[Wuzhen, China]GN[game2]
;B[qp];W[pd];B[cq];W[cd];B[ec];W[oq];B[pn];W[df];B[nc];W[qf]
;B[pc];W[qc];B[qb];W[oc];B[pb];W[od];B[ob];W[rc];B[nd];W[mb]
;B[lc];W[lb];B[qd];W[rd];B[jc];W[mc];B[pe];W[oe];B[ld];W[kp]
;B[iq];W[fp];B[dn];W[io];B[ch];W[cl];B[eh];W[cg];B[bg];W[bf]
;B[fn];W[em];B[en];W[ek];B[kq];W[lq];B[lp];W[jq];B[kr];W[jp]
;B[jr];W[mq];B[ip];W[ho];B[gp];W[dq];B[go];W[cp];B[lo];W[kn]
;B[ln];W[km];B[dp];W[hp];B[hq];W[gq];B[gr];W[fq];B[hn];W[jo]
;B[fr];W[do];B[co];W[ep];B[bp];W[md];B[ne];W[of];B[lm];W[le]
;B[kl];W[jl];B[gl];W[kk];B[ll];W[jk];B[mj];W[gj];B[pq];W[pr]
;B[qr];W[po];B[qo];W[on];B[op];W[np];B[pm];W[om];B[no];W[bo]
;B[dp];W[oo];B[pp];W[cp];B[cn];W[nl];B[lk];W[mo];B[pl];W[nj]
;B[ni];W[ok];B[nf];W[mi];B[li];W[mh];B[og];W[pf];B[ki];W[ij]
;B[lg];W[nh];B[oi];W[oh];B[pj];W[ph];B[pk];W[bq];B[dp];W[jg]
;B[kg];W[cp];B[jn];W[in];B[dp];W[fo];B[cp];W[gn];B[ke];W[me]
;B[jf];W[jb];B[ic];W[ib];B[er];W[hc];B[cc];W[bc];B[dd];W[cb]
;B[ce];W[dc];B[cf];W[dg];B[be])

Hiroshi Yamashita

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

[Computer-go] AlphaGo Retires

[Petr Baudis]

  After winning against Ke Jie 3-0

https://deepmind.com/research/alphago/alphago-china/

AlphaGo retires from competitive Go.  The team released 50 selfplay
games, will work on a teaching tool and publishing a second paper.

https://blog.google/topics/google-asia/alphagos-next-move/

https://deepmind.com/research/alphago/alphago-vs-alphago-self-play-games/

  Congratulations and thank you, AlphaGo team, for your open scientific
approach and publishing your methods.  The Computer Go community learned
a lot as well, as evident from nowadays high-ranked KGS bots.

-- 
Petr Baudis, Rossum.ai
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

[Computer-go] Mailing list working?

[J. van der Steen]

Hi all,

Is there something wrong with the mailing list? I didn't see any 
messages since the 23rd of May.


best regards,
Jan van der Steen
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

[Computer-go] Query

[Michael Alford]

I have seen no posts to the group since May 23, not even my own. What is 
happening? Could someone email me, please?



Michael

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

[Computer-go] Xeon Phi result

[Gian-Carlo Pascutto]

Hi all,

I managed to get a benchmark off of a Intel® Xeon Phi™ Processor 7250
16GB, 1.40 GHz, 68 core (272 thread) system.

I used a version of Leela essentially identical to the public Leela
0.10.0, but compiled with -march=knl (using gcc 5.3), using an
appropriate version of Intel MKL (2017.1 for MIC) and increasing the
maximum amount of threads.

benchmark:

~ 151000 g/s (557 g/s per thread)

netbench:

predictions ->   670 p/s
evaluations ->  3007 p/s

This was with the 16G HBM bound as addressable memory. Using the regular
DDR4 cuts "netbench" numbers in half, but has no big impact on "benchmark".

This means it's about 5 times faster (in integer operations) than a quad
core desktop with HT, and about similar in floating point performance to
a mid-range videocard.

It's a nice machine, if a bit pricey.

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

Re: [Computer-go] mini-max with Policy and Value network

[Gian-Carlo Pascutto]

On 24-05-17 05:33, "Ingo Althöfer" wrote:
>> So, 0.001% probability. Demis commented that Lee Sedol's winning move in
>> game 4 was a one in 10 000 move. This is a 1 in 100 000 move.
> 
> In Summer 2016 I checked the games of AlphaGo vs Lee Sedol
> with repeated runs of CrazyStone DL:
> In 3 of 20 runs the program selected P10. It
> turned out that a rather early "switch" in the search was
> necessary to arrive at P10. But if CS did that it
> remained with this candidate.

I guess it's possible this move is selected by a policy other than the
neural network. Or perhaps the probability can be much higher with a
differently trained policy net.

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

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

[Cai Gengyang]

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] Human Bias for 50 game selection? 0.5 Komi?

[David Ongaro]

Dear Aya,

I hope you still read this list and will answer a question once in a while even 
though AlphaGo has retired. On fb you answered Cho Seok-bins question "I heard 
that alphaGo master is 3stone stronger than sedol version.
Is it only rating deference? or did they have played with 3stone hadicap game?” 
with:

They played 3-stone handicap games with komi 0.5.

So I wonder how manage you overcome hardwired 7.5 Komi of the value network?

Another question floating around is how big the pool of games was from were the 
published 50 AlphaGo vs AlphaGo were selected from? And what was the selection 
criteria? Demis Hassabis mentioned that these were played with long time 
settings, so these don’t seem to be normal training games. Can you reveal what 
the actual time settings where? Were dedicated machines used for each AlphaGo?

Even if you can’t answer all question every answer is appreciated.

Thanks so much

David O.

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

[Computer-go] successor of UEC Cup will be held in December 9, 2017

[Hiroshi Yamashita]

Web page says,

-
The successor tournament of the UEC Cup will be held on the following schedule.
　Date: 9th to 10th December 2017
The organizer of this new event is "IGO & SHOGI CHANNEL INC.".
The regulation and rule of the competition will be planned to inherit past 
UEC-cup.
Details such as the venue will be announced in July.
(June 06, 2017)
http://www.computer-go.jp/uec/public_html/eng/index.shtml
-

Thanks,
Hiroshi Yamashita

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

Re: [Computer-go] Functioning?

[kevin long]

Maybe we are all depressed about it. 

On Jun 7, 2017 10:48 AM, "Michael Alford"  wrote:

> I haven't seen a post to the group in three days.  That can't be possible,
> with the event in China.
> ___
> 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] Multi-Labelled Value Networks for Computer Go

[Hiroshi Yamashita]

Hi,

CGI team published interesting paper.
Their value network is +130 Elo stronger than simple value network.

Multi-Labelled Value Networks for Computer Go
https://arxiv.org/abs/1705.10701

By the way, I had not received any mails since May 23.
And today I got three Kato's mail that posted May 24.
From archives, Jan van der Steen also reported same trouble.
http://computer-go.org/pipermail/computer-go/2017-May/010125.html

Thanks,
Hiroshi Yamashita

___
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

[Xavier Combelle]

Le 7 juin 2017 22:21, "Cai Gengyang"  a écrit :

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 ?


I would say together with the 60-0 win of the Master games yes.


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 ?


Yes it is the easiest way to make a strong bot.


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] Mailing list working?

[Erik van der Werf]

Yup, looks like something broke. Here everything that was sent after the
23rd only arrived today (June 7)... Ah well, it's game-over anyway :-)

On Mon, May 29, 2017 at 7:51 AM, J. van der Steen <
j.van.der.st...@gobase.org> wrote:

>
> Hi all,
>
> Is there something wrong with the mailing list? I didn't see any messages
> since the 23rd of May.
>
> best regards,
> Jan van der Steen
> ___
> 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] Mailing list working?

[hamburg.de]

So Computer-go achieved its goal and we can close this list?


> On Jun 7, 2017, at 15:33, Erik van der Werf  wrote:
> 
> Yup, looks like something broke. Here everything that was sent after the 23rd 
> only arrived today (June 7)... Ah well, it's game-over anyway :-)
> 
> On Mon, May 29, 2017 at 7:51 AM, J. van der Steen  > wrote:
> 
> Hi all,
> 
> Is there something wrong with the mailing list? I didn't see any messages 
> since the 23rd of May.
> 
> best regards,
> Jan van der Steen
> ___
> 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