[mlpack] Rotate new data with Kernel PCA

[Ryan Curtin]

On Wed, Apr 02, 2014 at 11:38:31PM +0000, dslate at speakeasy.net wrote:
> Hi Ryan,
> 
> Thanks for your answer.  I figured it should be possible.  I'm not very
> experienced with Support Vector Machines, but wouldn't they have to do
> this sort of thing to predict for new data?

Yes.  SVMs aren't my cup of tea, but they must be trained first, and (at
least some of) the training data must be retained for classification.
The reason mlpack's kernel PCA implementation doesn't have the
functionality you were looking for is that kernel PCA is not always used
in the train/test setting; sometimes it's just used to get a nonlinear
mapping for the training points, and there are no testing points.  That
was the scenario we considered when we wrote it...

> P.S. Good quote in your .sig from "Dr. Strangelove", a classic film.

Thanks.  I have a collection of quotes I found entertaining from various
movies, and they get randomly selected when I write an email.  Kubrick
movies are among my favorites.

-- 
Ryan Curtin    | "Lots of respectable people have been hit by
ryan at ratml.org | trains."  - Penny