I should be more specific. I don't mean looking at the results of global UCT to 
determine life and death. I mean a routine that gets called to test a single 
group. It has some nice properties, such as giving a probability even when it 
can't finish within the time allotted. Most of the tricky design decisions are 
ones which apply no matter what algorithm you use, i.e. when to break off an 
attack, and what policy to use to constrain moves to the "relevant" ones. 
 
One big difference wrt global UCT is the frequency with which it hits terminal 
nodes. I don't treat terminal nodes in any special way in global UCT. That 
would be a bad idea in life and death search.
 
- Dave Hillis 
 
-----Original Message-----
From: [EMAIL PROTECTED]
To: computer-go@computer-go.org
Sent: Mon, 16 Apr 2007 11:32 AM
Subject: Re: [computer-go] Using Monte Carlo Go/UCT for subgames


I think I remember that post - with the same level of clarity unfortunately. 
 
I use UCT for life and death. It comes into play at the scoring phase and in 
"polite" games where the engine needs to know when it can pass. (Knowing the 
score doesn't necessarily mean one knows the status of every stone.) I am happy 
with how it works and the level of effort it took to get there. But I haven't 
done any reasonable comparison to test how fast it is compared to other 
algorithms.
 
- Dave Hillis
 
 
-----Original Message-----
From: [EMAIL PROTECTED]
To: computer-go@computer-go.org
Sent: Mon, 16 Apr 2007 8:26 AM
Subject: [computer-go] Using Monte Carlo Go/UCT for subgames


Hi, I distantly remember a post, where someone said something 
about someone (sorry... I know this is really vaque) working 
on using Monte Carlo Go/UCT to estimate results in subgames, 
i.e., life & death, connection, capture, etc. 
 
Does anyone know, if there are any published results on 
the subject? 
 
-Esa 
 
_______________________________________________ 
computer-go mailing list 
computer-go@computer-go.org 
http://www.computer-go.org/mailman/listinfo/computer-go/ 



Check Out the new free AIM(R) Mail -- 2 GB of storage and industry-leading spam 
and email virus protection.

_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
________________________________________________________________________
Check Out the new free AIM(R) Mail -- 2 GB of storage and industry-leading spam 
and email virus protection.
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to