It is currently Fri Apr 26, 2024 8:06 pm

All times are UTC - 8 hours [ DST ]




Post new topic Reply to topic  [ 7 posts ] 
Author Message
Offline
 Post subject: Databases again
Post #1 Posted: Tue Mar 22, 2016 1:39 pm 
Lives in gote

Posts: 677
Liked others: 6
Was liked: 31
KGS: 2d
Obviously deep learning needs deep databases of good quality games (better than 5dKGS/8dTygem I guess). Anyone a clou where to find those or where those programmers get them? I could need a database with 1.000.000 quality games :).

Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #2 Posted: Tue Mar 22, 2016 1:49 pm 
Tengen

Posts: 4380
Location: North Carolina
Liked others: 499
Was liked: 733
Rank: AGA 3k
GD Posts: 65
OGS: Hyperpape 4k
I was about to say it's not obvious, since Google used one version of the neural net to play games which were then fed back into their system for reinforcement learning.

But if there was an open database of sufficient size, you could avoid the step of recreating games while training your system. Does anyone know how much of the resources that accounts for?

_________________
Occupy Babel!

Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #3 Posted: Thu Mar 24, 2016 6:00 pm 
Lives in gote
User avatar

Posts: 392
Liked others: 23
Was liked: 43
Rank: NR
deep learning does not need deep databases because if you already have a deep database you don't need to deep learn


Last edited by djhbrown on Tue May 02, 2017 1:07 am, edited 1 time in total.
Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #4 Posted: Fri Mar 25, 2016 4:56 am 
Lives in gote

Posts: 418
Liked others: 9
Was liked: 83
Rank: kgs 5 kyu
KGS: Pio2001
In the paper published by Deepmind in Nature, they say

Quote:
We trained the policy network (...) to classify positions according to expert moves played in the KGS data set. This data set contains 29.4 million positions from 160,000 games played by KGS 6 to 9 dan human players; 35.4% of the games are handicap games.


In the press conference after match 5, they said that Demis thought that AlphaGo could learn the game by itself, but they all the same started the deep learning with a database of human games.

I think that it would be interesting to see what fuseki would AlphaGo play if it had not seen any human game at all. But I'm afraid that it would miss some important concepts, like a human repeating the same mistakes against a weak software because the mistakes are never punished.
As they said in the interviews, neural networks have "blind spots". The more they are trained, the smaller the blind spots.

Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #5 Posted: Fri Mar 25, 2016 6:59 am 
Lives in gote

Posts: 502
Liked others: 1
Was liked: 153
Rank: KGS 2k
GD Posts: 100
KGS: Tryss
djhbrown wrote:
but since pros are only human (cogito ergo err) they make mistakes, and a database full of mistakes is no use to man or beast.

:roll:

You don't need perfect play for a database to be usefull... I even think that a database of games devoid of of mistakes would be less usefull


By the way, your joseki database is probably full of mistakes (the concept of joseki, is, in the absolute, a mistake. But in practice, at human level, it's verery usefull)

Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #6 Posted: Sat Apr 02, 2016 5:46 pm 
Lives in gote
User avatar

Posts: 392
Liked others: 23
Was liked: 43
Rank: NR
Tryss wrote:
the concept of joseki, is, in the absolute, a mistake.


Last edited by djhbrown on Tue May 02, 2017 1:08 am, edited 1 time in total.
Top
 Profile  
 
Offline
 Post subject: Re: Databases again
Post #7 Posted: Sat Apr 02, 2016 6:03 pm 
Oza

Posts: 2180
Location: ʍoquıɐɹ ǝɥʇ ɹǝʌo 'ǝɹǝɥʍǝɯos
Liked others: 237
Was liked: 662
Rank: AGA 5d
GD Posts: 4312
Online playing schedule: Every tenth February 29th from 20:00-20:01 (if time permits)
djhbrown wrote:
Tryss wrote:
the concept of joseki, is, in the absolute, a mistake.


oh, really? :) fair play is a mistake, eh? ...


Why do you say that? Fair play is a mistake to any political party in any country in the world. If it weren't they would never get elected.

_________________
Still officially AGA 5d but I play so irregularly these days that I am probably only 3d or 4d over the board (but hopefully still 5d in terms of knowledge, theory and the ability to contribute).

Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 7 posts ] 

All times are UTC - 8 hours [ DST ]


Who is online

Users browsing this forum: Bing [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group