nlopt - setting step size for derivative based optimization?

Question

I am trying to solve to following optimization problem:

1. minimize convex score function (quadratic) with respect to the optimization variables. with a few nonlin