Software

Feel free to distribute and modify the software in this page, but please maintain the credits.

A quadratic program solver

When we attempt to solve the Basis Pursuit (BP) problem



where is the L1 norm and is an underdetermined linear system of equations, in a scenario where the rows of are distributed, the following problem usually arises.



Note that the linear system in this problem contains just a few equations of in the BP, but the size of the variable is the same. We provide an efficient Matlab solver for this problem here. The solver uses the barrier method (an interior point algorithm) together with Newton's method, where the inversion of the hessian matrix is almost done in closed form, specially for the case where consists of just one equation. A more precise description of the method can be found here.