About the fact that ladders appeared so late:
- The learning was based on self-play. Understanding ladders is perhaps not
so important if your opponent doesn't understand them either... Every time
a decisive ladder appears on the board, the result is practically a coin
toss.
- And as others have
I saw my first AlphaGo Zero joke today:
After a few more months of self-play the games might look like this:
AlphaGo Zero Black - move 1
AlphaGo Zero White - resigns
Cheers,
David G Doshay
ddos...@mac.com
___
Computer-go mailing list
I also expected bootstrapping by self-play. (I also wrote a post to that
effect. But of course, DeepMind actually DID IT.)
But I didn't envision any of the other stuff. This is why I love their papers.
Papers from most sources are predictable, skimpy, and sketchy, but theirs
contain all sorts
Some thoughts toward the idea of general game-playing...
One aspect of Go is ideally suited for visual NN: strong locality of reference.
That is, stones affect stones that are nearby.
I wonder whether the late emergence of ladder understanding within AlphaGo Zero
is an artifact of the board
Yeah, I would expect that encoding stones as "signed liberty count" would train
faster/better/stronger. You could imagine a follow-up paper where Go features
are supplied.
But, if I think about this on a large scale, wouldn't it be huge to put
together a general game-playing program just based
On 18/10/2017 22:00, Brian Sheppard via Computer-go wrote:
> This paper is required reading. When I read this team’s papers, I think
> to myself “Wow, this is brilliant! And I think I see the next step.”
> When I read their next paper, they show me the next *three* steps.
Hmm, interesting way of
On 18/10/2017 22:00, Brian Sheppard via Computer-go wrote:
> A stunning result. The NN uses a standard vision architecture (no Go
> adaptation beyond what is necessary to represent the game state).
The paper says that Master (4858 rating) uses Go specific features,
initialized by SL, and the same
On 18/10/2017 19:50, cazen...@ai.univ-paris8.fr wrote:
>
> https://deepmind.com/blog/
>
> http://www.nature.com/nature/index.html
Select quotes that I find interesting from a brief skim:
1) Using a residual network was more accurate, achieved lower error, and
improved performance in AlphaGo by
This paper is required reading. When I read this team’s papers, I think to
myself “Wow, this is brilliant! And I think I see the next step.” When I read
their next paper, they show me the next *three* steps. I can’t say enough good
things about the quality of the work.
A stunning result.
A link to the paper (from the blog post):
https://deepmind.com/documents/119/agz_unformatted_nature.pdf
Enjoy!
Álvaro.
On Wed, Oct 18, 2017 at 2:29 PM, Richard Lorentz
wrote:
> Wow! That's very exciting. I'm glad they didn't completely shelve the
> project as they
They are 80 games of different version of alphago and 3 of alphago
against same version of alphago in supplementary data of
https://www.nature.com/nature/journal/v550/n7676/full/nature24270.html#supplementary-information
Le 18/10/2017 à 20:29, Richard Lorentz a écrit :
> Wow! That's very
Wow! That's very exciting. I'm glad they didn't completely shelve the
project as they implied they might do after the match with Lee Sedol.
I'm looking forward to seeing some games and "... plays unknown to
humans", as Hassabis states.
Also, I love this comment from Silver, something I have
https://deepmind.com/blog/
http://www.nature.com/nature/index.html
Impressive!
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go
13 matches
Mail list logo