[Eeglablist] AMICA - Using PCA parameter

[Jason Palmer]

Hi Cortney,

 

The sphering matrix used in the dimensionality reduction case is the first
num_pcs rows of the sphering matrix.

 

So the ICA unixing matrix would be

 

                >> numpcs = 150;

                >> activations  = weights * sphere(1:numpcs,:) * (input_data
- row_means);

 

Best,

Jason

 

 

From: eeglablist-bounces at sccn.ucsd.edu
[mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Cortney Bradford
Sent: Tuesday, March 12, 2013 12:36 PM
To: eeglablist at sccn.ucsd.edu
Subject: [Eeglablist] AMICA - Using PCA parameter

 

We are interested in using using the PCA parameter when running AMICA
(amica11).

 

sample of the parameter settings:

 

data_dim 229

doPCA 1

pcakeep 150

 

 

AMICA seemed to run fine, but produced an ICA weights matrix that is 150x150
and sphere matrix that is 229x229. This becomes a problem when we try to
compute the activations using the following equation, due to the matrices
being different sizes. 

 

activations = weights*sphere*(input_data - row_means);



Are we doing something wrong in setting parameters or is this a bug?
 
Thank you,
 
J. Cortney Bradford
Post-doc
 
University of Michigan






 
 
 
 
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130312/3d5a2fc4/attachment.html>