Path: news.mathworks.com!newsfeed-00.mathworks.com!newsfeed2.dallas1.level3.net!news.level3.com!postnews.google.com!e18g2000vbe.googlegroups.com!not-for-mail
From: Rune Allnor <allnor@tele.ntnu.no>
Newsgroups: comp.soft-sys.matlab
Subject: Re: how to deal with the inversion problem of a huge sparse
Date: Sun, 27 Sep 2009 06:32:49 -0700 (PDT)
Organization: http://groups.google.com
Lines: 19
Message-ID: <56a8acfa-3493-43d8-94bf-76a7c9b3b115@e18g2000vbe.googlegroups.com>
References: <h9llp6$mho$1@fred.mathworks.com> <ace8cc81-2737-4630-83b6-76c09f427a97@p23g2000vbl.googlegroups.com> 
	<524bfbe2-ff4e-4cc6-a2d0-aa05d821fbd1@l35g2000vba.googlegroups.com> 
	<h9nnq6$khh$1@fred.mathworks.com>
NNTP-Posting-Host: 77.16.41.21
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
X-Trace: posting.google.com 1254058370 592 127.0.0.1 (27 Sep 2009 13:32:50 GMT)
X-Complaints-To: groups-abuse@google.com
NNTP-Posting-Date: Sun, 27 Sep 2009 13:32:50 +0000 (UTC)
Complaints-To: groups-abuse@google.com
Injection-Info: e18g2000vbe.googlegroups.com; posting-host=77.16.41.21; 
	posting-account=VAp5gAkAAAAmkCze5hvZtMeedpZWNthI
User-Agent: G2/1.0
X-HTTP-UserAgent: Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 5.1; 
	Trident/4.0; .NET CLR 2.0.50727; .NET CLR 3.0.04506.648; .NET CLR 3.5.21022; 
	.NET CLR 3.0.4506.2152; .NET CLR 3.5.30729; .NET CLR 1.1.4322),gzip(gfe),gzip(gfe)
Xref: news.mathworks.com comp.soft-sys.matlab:573190


On 27 Sep, 15:03, "Bruno Luong" <b.lu...@fogale.findmycountry> wrote:
> Rune Allnor <all...@tele.ntnu.no> wrote in message <524bfbe2-ff4e-4cc6-a2d0-aa05d821f...@l35g2000vba.googlegroups.com>...
>
> > Time has nothing to do with the statistical properties.
> > Assign any physical dimension you want to the data. I still
> > can't see why one would need large sparse correlation matrices.
>
> A very simple example is all the data are all independent. The correlation matrix is an identity matrix. May be you don't consider an identity matrix as sparse? Or you think there is never two data independent?

Assume you have, say, a 1,000,000x1 data vector with independent
samples. If one takes your argument ilterally, you seem to say that
one needs a 1,000,000x1,000,000 identity matrix to represent
the covariance.

My argument is that you only need a small portion of that, say,
a 10x10 matrix. You have the same information but in a far more
convenient form.

Rune