principal components analysis

[falkpl]

Has anyone done Principal Components Analysis in labview?  Essentially I need it for patern recognition in n dimensions (8-14 dimensional feature set space)  this will involve doing a covariance on a mxn matrix and then getting the eigenvectors/eigenvalues.  I am not a math wiz and avoid such analysis when possible but have to use PAC for a particular problem.  I am using LV 7.0 right now.  Anyone have code for such a problem?

Paul

[Travis_M.]

The Advanced Signal Processing Toolkit addresses PCA in the "Time Series Analysis Tools." The following is the product page for that toolkit, where if you click the link to "Manuals" in the "Resources" section (highlighted in blue) you'll find the "Time Series Analysis Tools User Manual" linked:

NI LabVIEW Advanced Signal Processing Toolkit
http://sine.ni.com/nips/cds/view/p/lang/en/nid/1395

For convenience, here is the link to the manual directly:

http://digital.ni.com/manuals.nsf/websearch/EE315092E16D3AC08625703F002356EC

Also, here is the toolkit's release document on-line:

http://digital.ni.com/worldwide/bwcontent.nsf/websearch/286eb6b494fc4a578625704a00747a85

I don't know about its particular use with your application, but this might help!

[falkpl]

Thanks for the links, I will definatly look them over.  PCA after reading up on it is not all too bad.  Take the mxn 2d Data array, find the covarience get the eigen vectors, sort them by eigenvalue and then apply the principle component.  I am not a big stats fan (ironic for a computer science guy) but I cant avoid this one.  Thanks again.

Paul

[tst]

CC did some PCA on the list a while back. It might help you.