Meeting with Gert

I had a meeting with Gert this afternoon. We focused on the dual formulation of the sparse CCA problem. The sparse CCA problem can viewed as a quadratically constrained quadratic program (QCQP), (one is optimizing a quadratic function xX'Y'y over the vectors x,y, subject to the quadratic constraints x'X'Xx = 1 and y'Y'Y'y = 1). This program can be written into a semidefinite program (SDP) and the dual can be derived (which is also an SDP).

My dual formulation is essentially correct. Though there are some modifications that Gert and I spoke of that I will need to include regarding how the cardinality constraints are introduced. So that means that I will be going back to the pencil and paper and crunching out a little bit more math.

Right now I need to get some data off our servers and onto my lap top because i'm going to be out of town the next 2 weeks and I'd like to start some demo experiments with our audio data. I'm pretty sure I can get at least a good 4 hours of work in each day. Most of my friends back in town are back in school, so its not like i'll be very distracted on weekdays.