So that could affect the triangulation, maximal clique size etc? Seems like
one would want the optimization to trade off accuracy and updating
efficiency.
Bob
----- Original Message -----
From: Alex Kozlov <alexvk@vostok.engr.sgi.com>
To: Bob Welch <bwelch@gensym.com>
Cc: <uai@CS.ORST.EDU>; <alexvk@vostok.engr.sgi.com>
Sent: Friday, April 07, 2000 4:58 PM
Subject: Re: [UAI] mixed variables
> Bob Welch wrote:
> >
> > What is the constraint that prevents the number of discrete cells going
to
> > infinity?
> >
>
> There is a parameter $\delta$ in the model (see Fig. 6.12 on page 112).
>
> In short, the discretizations stops when an estimate of the
> discretization
> for a particular clique is smaller then $\delta$. The estimatate is not
> an exact upper bound, but works well enough.
>
> --
> Alexander V. Kozlov | alexvk@engr.sgi.com | (650) 933-8493
This archive was generated by hypermail 2b29 : Sat Apr 08 2000 - 08:32:13 PDT