[gnugo-devel] some logistical stuff...

[Eric]

--- Eric <address@hidden> wrote:

> In 1995, we learned how to make brute force more
> efficient: instead of immediately starting to build
> a
> search tree, the program first builds a
> polynomial-sized graph. Second, the program searches
> the graph. So there are 2 phases, and the program's
> execution switches back and forth between the
> 'build'
> phase and the 'search' phase.
> 
> The points here are that some automatic inferencing
> is
> being done during the build phase, and that the
> build
> phase inferences will subsequently be used to guide
> the search phase.
> 
> Further, since the graph is polynomial in size, and
> since it takes polynomial time to build the graph,
> then the inferences are coming to us at a cheap
> cost.
> They are being computed *outside* of the search
> space.
> Because search is too expensive.
> 
> This was the lesson of the paper:
> 
> Blum, A. and Furst, M.L. 1995, "Fast Planning
> Through
> Planning Graph Analysis". In Proc. International
> Joint
> Conference on Artificial Intelligence, pp.
> 1636-1642.
> ---------------

Hey, in thinking about the Graphplan planner folks,
their planner would work very nicely in the GNU Go
situation.

If we just happen to stumble across some planner
source code on the Web, are we allowed to bundle and
distribute it along with GNU Go? ... or?

Maybe here is the answer to a logistical problem:

http://www-2.cs.cmu.edu/People/avrim/graphplan.html