Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 0 down vote favorite

I have a question about Principal Components Analysis. I've tried to read a lot about it but have a doubt about this application.

I have high dimensional data (known to be similar) which is organized into the columns of matrix Q. I perform a princomp on this data, and project the centered columns of Q onto the calculated eigenvector basis (the first 4).

My question is, can I use the linearly independent vectors of the result A to form a basis for which I can test other high dim data for similarity? e.g. Solve Ax=b for an exact solution? I think the usual way of determining similarity is by Euclidean distance, but I wonder if this is possible too.

share|improve this question
1  
Just to clarify - what do you mean by "known to be similar"? (OT: Interesting username - is it novel or derived from something?) – Iterator Aug 15 '11 at 1:41
It's definitely possible to compare two datasets, however are you looking for a statistical formulation of a test, or more of a descriptive formulation (e.g. the difference between a t-test and Euclidean distance)? Both are feasible, but are very different questions. – Iterator Aug 15 '11 at 1:44
The data is comprised of training images that have similar parameters. I would like to get other (test) images that "fit into the group". Thus, the parameters are unknown for the test images and I thought I could use principal component analysis. The username is from the subreddit fifthworldproblems. – FiFThWoRlDFreaKo Aug 15 '11 at 1:48
Is this question, also on images, similar to yours? stackoverflow.com/questions/7061004/… – Iterator Aug 15 '11 at 1:50
It's a very different methodology, certainly. – Iterator Aug 15 '11 at 1:51
show 1 more comment

Know someone who can answer? Share a link to this question via email, Google+, Twitter, or Facebook.

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.