So, I read Rohklin and Tygert’s paper about using randomized preconditioning to solve overdetermined least squares problems. Their idea is, given a full rank matrix , where , and a vector , ins...

The “Givens rotation problem” is stuck in the back of my head. Namely, randomly subsampled Fourier matrices seem to serve excellently as random projections which preserve all sorts of nice...

The main reason I occasionally choose Matlab over Mathematica nowadays is the CVX library, which provides a friendly and powerful interface for convex programming. Unfortunately, Mathematica doesnR...

Here’s an interesting problem that my office mate described to me; it’s motivated by a particular way of modeling communication networks. Let be independent Bernoulli random variables (an...

I implemented the Projected Gradient Quasi-Newton method (PQN-NNLS) for solving the nonnegative least squares problem in Matlab, to compare to the idea I’m working on for a ‘fast’ in...