We consider a class of nonlinear optimization problems with "expensive" objective function and constraints. For such problems the derivatives of the objective function and, possibly, constraints are assumed unavailable. Also some noise may be present in the function computations. We will describe a class of, so called, derivative free optimization (DFO) algorithms developed for constrained and unconstrained problems.
This is joint work with Katya Scheinberg and Phillipe Toint.
