Re: About Principle Components Analysis
- From: Horand.Gassmann@xxxxxxxxxxxxxx
- Date: Mon, 18 Feb 2008 05:05:42 -0800 (PST)
On Feb 17, 12:36 pm, dtian.ty...@xxxxxxxxxxxxxx wrote:
Hello,
I am a little confused about the projection operation of PCA. It is
defined as E*X where E is an eigenvector; X is a data point and * is
the matrix multiplication operator. In the linear algebra textbooks
such as "Linear Algebra (Schaum's Outlines)" by Seymour Lipschutz,
projection of a vector u on a vector v is
[(u dotproduct v)/(square of magnitude of v)] v
If E is an eigenvector, then so is tE for any (nonzero) scalar t. It
is customary to normalize E, although there are several ways this is
done. If you use the L-2 norm, you get the simplified formula. Other
normalizations I have seen: L-1 norm; L-oo norm.
(By the way, you should spell it "Principal Component Analysis"...)
.
- References:
- About Principle Components Analysis
- From: dtian . tyson
- About Principle Components Analysis
- Prev by Date: Re: About Principle Components Analysis
- Next by Date: Re: How to do 4PL Logistic Regression
- Previous by thread: Re: About Principle Components Analysis
- Index(es):
Relevant Pages
|
