Re: Polygames: Improved Zero Learning

--===============4714785932472974559==
Content-Type: multipart/alternative; boundary="0000000000001bfca4059d984ed7"

--0000000000001bfca4059d984ed7
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Yep, global pooling is also how KataGo already handles multiple board sizes
with a single model. Convolution weights don't care about board size at all
since the filter dimensions have no dependence on it, so the only issue is
if you're using things like fully-connected layers, and those are often
easily replaceable with global pooling and convolution, making the whole
net board-size-independent. (Although you still need to train/finetune
appropriately).

Nice to see the same idea finally making its way around in papers too.
Along with all the experiments that projects like LC0 and such are trying
in different runs, it feels like the state of formal published research in
this area is often one step behind what major projects are already
successfully doing.

On Sun, Feb 2, 2020 at 8:06 AM R=C3=A9mi Coulom <[email protected]> wro=
te:

> Hi,
>
> I have just noticed this has recently been released:
>
> github:
> https://github.com/facebookincubator/polygames
> paper:
> https://arxiv.org/abs/2001.09832
>
> R=C3=A9mi
> _______________________________________________
> Computer-go mailing list
> [email protected]
> http://computer-go.org/mailman/listinfo/computer-go
>

--0000000000001bfca4059d984ed7
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Yep, global pooling is also how KataGo already handles mul=
tiple board sizes with a single model. Convolution weights don&#39;t care a=
bout board size at all since the filter dimensions have no dependence on it=
, so the only issue is if you&#39;re using things like fully-connected laye=
rs, and those are often easily replaceable=C2=A0with global pooling and con=
volution, making the whole net board-size-independent. (Although you still =
need to train/finetune appropriately).<div><br></div><div>Nice to see the s=
ame idea finally making its way around in papers too. Along with all the ex=
periments that projects like LC0 and such are trying in different runs, it =
feels like the state of formal published research in this area is often one=
 step behind what major projects are already successfully doing.</div></div=
><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Su=
n, Feb 2, 2020 at 8:06 AM R=C3=A9mi Coulom &lt;<a href=3D"mailto:remi.coulo=
[email protected]">[email protected]</a>&gt; wrote:<br></div><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div>Hi,</div><div><br><=
/div><div>I have just noticed this has recently been released:</div><div><b=
r></div><div>github:<br></div><div><a href=3D"https://github.com/facebookin=
cubator/polygames" target=3D"_blank">https://github.com/facebookincubator/p=
olygames</a></div><div>paper:</div><div><a href=3D"https://arxiv.org/abs/20=
01.09832" target=3D"_blank">https://arxiv.org/abs/2001.09832</a></div><div>=
<br></div><div>R=C3=A9mi<br></div></div>
_______________________________________________<br>
Computer-go mailing list<br>
<a href=3D"mailto:[email protected]" target=3D"_blank">Computer-g=
[email protected]</a><br>
<a href=3D"http://computer-go.org/mailman/listinfo/computer-go" rel=3D"nore=
ferrer" target=3D"_blank">http://computer-go.org/mailman/listinfo/computer-=
go</a><br>
</blockquote></div>

--0000000000001bfca4059d984ed7--

--===============4714785932472974559==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Computer-go mailing list
[email protected]
http://computer-go.org/mailman/listinfo/computer-go

--===============4714785932472974559==--