Re: Branch & bound for not everywhere defined constraints
Please see my inserted comments.
Baker
Dan Zuras Intervals wrote:
Date: Thu, 3 Sep 2009 12:53:16 -0400
From: Nate Hayes <nh@xxxxxxxxxxxxxxxxx>
Subject: Re: Branch & bound for not everywhere defiend constraints
To: STDS-1788@xxxxxxxxxxxxxxxxx
.
.
.
As I said, no computers were harmed in this process
so even this order may have problems.
What do you think now? Am I there yet?
That's what I get for wetware compiling. :-)
From what I gather in this mostly 3-way exchange between
Nate, Arnold, and Dan, there are various implementations of the
branch and bound algorithm depending upon whether we have
multiple NaI or not, and the main issue is one of efficiency.
First, I wish to thank Nate for posting his example of actual
computation. Second, I'm a bit confused about
(a) what is actually most efficient, and
(b) what is being proposed for multiple NaIs (that is, how
it is envisioned they might be implemented).
Along these lines, might it be worthwhile for someone to
volunteer to actually implement, somehow, the alternatives
in a way that they CAN be experimentally compared?
One issue which would seem to impact practicality of the
standard and simplicity of the specification is whether
the NaI (or NaIs) is envisioned as a global flag(s) or whether we mandate
it (them) as a part of the interval datum. Such things have
been argued here, but it isn't clear to me what the implications
of choosing one or the other of these alternatives is, other than
that part of the datum seems simpler to describe and specify in
a standard, at least at first glance. I'll need to rely on
others to shed more light on these things.
Baker
--
---------------------------------------------------------------
R. Baker Kearfott, rbk@xxxxxxxxxxxxx (337) 482-5346 (fax)
(337) 482-5270 (work) (337) 993-1827 (home)
URL: http://interval.louisiana.edu/kearfott.html
Department of Mathematics, University of Louisiana at Lafayette
(Room 217 Maxim D. Doucet Hall, 1403 Johnston Street)
Box 4-1010, Lafayette, LA 70504-1010, USA
---------------------------------------------------------------