I also wonder about this. A purely convolutional approach would save a lot of
weights. The output for pass can be set to be a single bias parameter,
connected to nothing. Setting pass to a constant might work, too. I don't
understand the reason for such a complication.
- Mail original
Seems like extraordinarily fast progress. Great to hear that.
-Original Message-
From: Computer-go [mailto:computer-go-boun...@computer-go.org] On Behalf Of
"Ingo Althöfer"
Sent: Friday, December 29, 2017 12:30 PM
To: computer-go@computer-go.org
Subject: [Computer-go] Project Leela Zero
As far as a purely convolutional approach, I think you *can* do better by
adding some global connectivity.
Generally speaking, there should be some value in global connectivity for
things like upweighting the probability of playing ko threats anywhere on
the board when there is an active ko
Hello in the round,
I am not sure how narrowly people from the list are
following the progress of Gian Carlo Pascutto's project
Zero Leela. Therefore, here are some impressions.
The project site is:
http://zero.sjeng.org/
Shortly before Christmas some guys in the German
Go mailing list claimed
I agree that having special knowledge for "pass" is not a big compromise, but
it would not meet the "zero knowledge" goal, no?
-Original Message-
From: Computer-go [mailto:computer-go-boun...@computer-go.org] On Behalf Of
Rémi Coulom
Sent: Friday, December 29, 2017 7:50 AM
To:
Seems suprisingly strong. Given that no super vcluster availab,´le for
trainning. Have at least on accoutn rated would be nice since in unrated
games people experiment quite a lot at cost of playing well.
2017-12-29 22:02 GMT+02:00 Brian Sheppard via Computer-go <
computer-go@computer-go.org>: