I'm trying to implement a paper I read and am having trouble setting it up. The goal is to learn a transformation matrix between two spaces containing the same set of labeled points.

So given a set of vector pairs alt text Where alt text

The transformation matrix W can be learned by optimizing
alt text

using stochastic gradient descent.

I'm having some trouble figuring out the gradient though. Any help or tips?

asked Oct 22 '14 at 11:12

pv123's gravatar image

pv123
1613

edited Oct 22 '14 at 13:39

Be the first one to answer this question!
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.