Re: PP vs. ICA

From: Jay Liu (liuj5_at_mcmaster.ca)
Date: 09/23/04 

Date: 23 Sep 2004 08:51:25 -0700

Thanks Aleks,

>From the viewpoint of application, the difference between ICA and PP
becomes clear. ICA is used to separate signal sources from mistures
blindly and PP is used for finding interestingness of multivariate
data in lower dimension (PP is rather exploratory).

However, from the viewpoint of a problem formulation, two are almost
identical. As you know we can formulate ICA and PP in terms of
optimization problems. Basis vectors are the optimization variables to
be determined. Non-gaussianity such as negative entropy can be used as
a performance index (or objective function) in both ICA and PP.

BTW, One difference I've noticed is that in ICA, PCA is often done
first to reduce the dimension when the number of sources is smaller
than the number of sensor measurements. But I haven't seen this in PP.

Can anyone share his/her idea with me?