Smoothed Functionals in Stochastic Optimization

We consider the problem of finding a global extremum for a nonconvex function by constructing another smoothed function, which has a better “behavior” than the original one. Then, operating only with the smoothed function, we find the global extremum for the original function.