The AlphaGo Zero paper tests 20 and 40 block resnet architectures (among
others), and in the AlphaZero paper AlphaZero Go plays against the 20-block
AlphaGo Zero, but I cannot find any mention of which architecture AlphaZero
is using! I'm assuming they are using either the 20 or 40 block resnets -
Thanks!
It would be interesting to see the performance of the policy network alone
in chess and shogi too.
There is no such plot in the arxiv paper.
Honestly, I don't expect it to be that good since engines without a
lookahead search never performed
that well in these domains -- unlike Go which