>I haven't tried it, but (with the computer chess hat on) these kind of
>proposals behave pretty badly when you get into situations where your
>evaluation is off and there are horizon effects.
In computer Go, this issue focuses on cases where the initial move ordering is
bad. It isn't so much
The tree built by MCTS is very unbalanced - some branches
are explored more thoroughly than others.
Tweaking the algorithm to favor the newer results might result
in an overall improvement, but it also would be subject to all
the pitfalls of partially or unevenly evaluated trees.
This is
The tree built by MCTS is very unbalanced - some branches
are explored more thoroughly than others.
Tweaking the algorithm to favor the newer results might result
in an overall improvement, but it also would be subject to all
the pitfalls of partially or unevenly evaluated trees.
This is
Cool, thanks.
On Mon, Jul 24, 2017 at 10:30 AM, Gian-Carlo Pascutto wrote:
> On 24-07-17 16:07, David Wu wrote:
> > Hmm. Why would discounting make things worse? Do you mean that you
> > want the top move to drop off slower (i.e. for the bot to take longer
> > to achieve the
On 24-07-17 16:07, David Wu wrote:
> Hmm. Why would discounting make things worse? Do you mean that you
> want the top move to drop off slower (i.e. for the bot to take longer
> to achieve the correct valuation of the top move) to give it "time"
> to search the other moves enough to find that
Thanks for the replies!
On Mon, Jul 24, 2017 at 9:30 AM, Gian-Carlo Pascutto wrote:
> On 23-07-17 18:24, David Wu wrote:
> > Has anyone tried this sort of idea before?
>
> I haven't tried it, but (with the computer chess hat on) these kind of
> proposals behave pretty badly when
Sent: Sunday, July 23, 2017 12:25 PM
To: computer-go@computer-go.org
Subject: [Computer-go] Possible idea - decay old simulations?
I've been using Leela 0.10.0 for analysis quite often, and I've noticed
something that might lead to an improvement for the search, and maybe also for
other MCTS
I've been using Leela 0.10.0 for analysis quite often, and I've noticed
something that might lead to an improvement for the search, and maybe also
for other MCTS programs.
Sometimes, after putting hundreds of thousands of simulations into a few
possible moves, Leela will appear to favor one,