NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/sci/math/numanal/2555

Internet Message Format | 1992-08-26 | 1.9 KB

Xref: sparky sci.math.num-analysis:2555 sci.math.stat:1756 sci.math:10599
Path: sparky!uunet!dtix!darwin.sura.net!jvnc.net!yale.edu!qt.cs.utexas.edu!cs.utexas.edu!uwm.edu!ux1.cso.uiuc.edu!roundup.crhc.uiuc.edu!uirvlh!hougen
From: [e-mail] (Darrell Roy Hougen)
Newsgroups: sci.math.num-analysis,sci.math.stat,sci.math
Subject: Re: Least Squares with Inequality Constraints?
Keywords: least squares, inequality constraints
Message-ID: <[e-mail]>
Date: 25 Aug 92 22:04:54 GMT
References: <[e-mail]> <[e-mail]>
Distribution: inet
Organization: Center for Reliable and High-Performance Computing, University of Illinois at Urbana-Champaign
Lines: 20
NNTP-Posting-Host: uirvlh.csl.uiuc.edu

As an addendum to my note, I should point out that I had a similar
question some time ago.  Several people sent responses which I later
posted.  Currently, I am using a nonnegative least squares program
provided to me by another person.  It has the desirable properties
that it usually finds a solution in a small number of iterations and I
think it even keeps the partial results from one iteration to the next
so that only one column/row of the matrix has to be updated on each
iteration.  It is pretty fast.

Unfortunately, the program I am using only handles the case where all
of the variables are required to be positive and I would like to let
some of the variables range over the reals.  I would really like to
get my hands on the algorithm of Stoer '71 as it appears to be very
good.  Is this available anywhere?  At this point, I don't care too
much whether it is free or not.  However, because I have a *very*
large number of equations (e.g 10,000) in a *very* large number of
unknowns (e.g. 30), the interface should be such that I don't have to
type in all the data by hand (as that would be impossible).

Darrell