Representation, results on defining a discussion goal

[Pieter Cuijpers <P.J.L.Cuijpers@xxxxxxxxxxx>]

Hello Folks !

So far I had 5 replies on my call for possible goals in a discussion
about a usefull representation of the game of Go.
I will try to give an independent summary of all
the contributions including my own and of a 
small search I did to find what was written about 
the subject in former e-mails on this list.
Also a recent remark by Serge Boisse seems
of intrest to this discussion to me.

Summaries:
(consisting of small quotes and [possibly wrong] interpretations)

Mika Kojo:
IMO the issue of representation should have a specific context.
[neural networks, frameworks for Go programs (what subprograms
does a good Go-P. consist of), live-dead, board position, local tactical search, global move selection]

P.J.Leonard:
What do you mean by "representation" ?

Heikki Levanto:
I can imagine that some models might be good for programming the game mechanics and different ones would be needed for strategic decisions.
[stones-strings-groups-territories-influence,
neural net for high level decisions,
prolog-like rule based system]

John Aspinall:
I think the best hope for smart computer go is multiple 
representations. The problem then will be mapping between
representations.
[singly rooted game tree
multiple trees representing sub-games,
abstracting the topology,
reasoning backwards,
hashcoding and other identifiers for board states]

Pieter Cuijpers:
All the people who answered to my call, I think agreed on one
subject: One representation can not do the job, we need different
representations for different goals.
This is quite different from the statement I read in Serge Boisse's
reply on Intelligence (06 May 1999 10:48 AM (dutch time?))

Serge Boisse:
<skipped>
So, even though adding knowledge to a go program could increase 
its ability, I tell you that this is not enough, and that no program will
ever be able of capturing the deep essence of a 19x19 game.
...
This is somewhat a disapointing statement. But I will stand to that
point unless some genious mathematician proves me that the game
has a yet cached structure (fractal ?) that enables understand the
grounds of the game a different way.

Pieter Cuijpers:
This reminds me of an exam in the "History of Technology" I am 
planning to do tomorow. There was a piece about Maxwell's
equations in electrical engineering.
The only thing that Maxwell realy did was to take four of the leading
theorems about electricity and magnetism and put them together
in one framework. This framework became the new representation
for electro-magnetic phenomena.
(So Maxwell was the mathematical genius in this case. I hope that
an effort of multiple people may result in a mathematically genius
representation for Go)

In a way this is what I am trying to reach.
But this approach has some (great) disadvantages at the moment.

1. Maybe not all the "sub-representations" are not known yet.
2. Those who are known (life-death, influence, etc.) are often
not full-grown and may therefore not  be suitable.
3. I'm not even 1 dan (hope to be soon but that's not improtant now)
and I need a lot of independent oppinions to get to a full understand
of what is important in the game of Go.

So far so good.
After reading quite a few old emails I only discovered the need
for good representations for neural net computing, knowledge 
based decission making, life and death etc. again.
No one has started a discussion yet on how to integrate those
ideas in one larger model.

I do not want to start a discussion on whether it is possible to 
create an integrated model for Go. I hope it's possible or let
my determination to do it make it possible ;-)
Also I do not want to start a discussion on what an integrated
model should contain because this is probably the same
question as "what should a good Go program contain ?"
as was asked many times on this mailing list I think.

My question is:
Tell me a (or the best) way to develop a representation or model
of Go that contains multiple (or all) elements of the game.

I know it's silly but be a philosopher instead of an engineer.

(As Nicolai Hell said in Trevanians "Shibumi":
Go is to chess what philosophy is to double input accounting.)

Greetz. Pieter Cuijpers

PS. Again pls. send the replies to me personally and I will try to
handle the data...