> I fixed allowing multiple suicide in playouts, but it didn't make the
> many wasted playouts go away on Sheppard's position.
> On further investivation, the problem has to do with the interaction
> between superko and the transposition table.
> Currently, Orego checks only simple ko most of t
In message <20090807092625.gj15...@vanheusden.com>, Folkert van Heusden
writes
What is superko?
My program keeps a list of all board-positions and then if it whants to
do a move it checks if the new board-position is in the list. If so, it
throws that move away. Are there other checks I need t
>> What is superko?
>> My program keeps a list of all board-positions and then if it whants to
>> do a move it checks if the new board-position is in the list. If so, it
>> throws that move away. Are there other checks I need to do as well?
>
> Superko involves repeating a previous board position.
>> What is superko?
>> My program keeps a list of all board-positions and then if it whants to
>> do a move it checks if the new board-position is in the list. If so, it
>> throws that move away. Are there other checks I need to do as well?
>
> Superko involves repeating a previous board position.
On Fri, Aug 7, 2009 at 5:51 AM, Folkert van Heusden
wrote:
> >> What is superko?
> >> My program keeps a list of all board-positions and then if it whants to
> >> do a move it checks if the new board-position is in the list. If so, it
> >> throws that move away. Are there other checks I need to do
I have traced many, many bad moves to RAVE failures. This post
is a "brain dump" on what I have learned.
SYMPTOMS
I classify two failure modes:
1) RAVE searches a bad move because it is good later.
2) RAVE won't search the best move because it is bad later.
The first failure mod
Perhaps the "context" attached to RAVE needs to be more subtle than a static
3x3 pattern - tactical and efficiency considerations may apply - a move may be
good when it defends or kills a group, but bad if it has no effect upon the
status - it may be wasted in such cases.
Terry McIntyre
The other alternative is to use domain knowledge as well as rave to bias
moves in the UCT search. I think Crazystone and Zen use learned patterns
for bias. Many Faces uses the MF knowledge engine combined with rave to
bias UCT search. I think this is stronger than trying to fix rave.
David
---
> -Original Message-
> From: terry mcintyre
> To: computer-go
> Sent: Fri, Aug 7, 2009 10:35 am
> Subject: Re: [computer-go] RAVE problems
>
> Perhaps the "context" attached to RAVE needs to be more subtle than a static
> 3x3 pattern - tactical and efficiency > considerations may app
I'll add some detail. I have 10 features which give me 1024 possible context
codes, not all of which are realizable. I keep a CAMAF table of approximate
size 1024*(number of spaces)*colors (1024*81*2 for a 9x9 board). This table
persists throughout the entire game, with suitable decays of the we
Although I've never done a RAVE implementation (soon, very soon), this is
related in the sense that AMAF is related to RAVE:
I have recently (yesterday, actually) done some experiments on AMAF with 5-vertex patterns (normal AMAF can be considered 1-vertex patterns). I was not able to
observe a
11 matches
Mail list logo
