Using Kernel-PCA with SVM

Is there any benefit to running a dimensionality reduction (e.g. Kernel PCA) before using the SVM algorithm? Naively, it seems like feeding Kernel-PCA-preprocessed data to a linear SVM is a similar process to feeding the raw data to kernel-SVM. I'm trying to develop some intuition about whether and how the two strategies differ.

asked Mar 26 '13 at 10:16

Chris Beaumont
31●2●2●5

There might be a benefit if you have unlabeled data, as that will implicitly perform some kind of semi supervised learning.

(Mar 26 '13 at 12:36) Alexandre Passos ♦

Subscription:

Once you sign in you will be able to subscribe for any updates here

Tags:

svm ×134
dimensionality-reduction ×31
kernel-trick ×10

Asked: Mar 26 '13 at 10:16

Seen: 627 times

Last updated: Mar 26 '13 at 12:36

Using Kernel-PCA with SVM

Subscription:

Related questions