Stochastic Perturbation Methods for Affine Restrictions

This work concerns the global optimization of a nonconvex objective function under linear restrictions. A suitable random perturbation of the projected gradient descent method leads to global convergence. Numerical examples are given.