Global optimization problems in which evaluation of the objective function is an expensive operation arise frequently in engineering, machine learning, decision making, statistics, optimal control, etc. A general global optimization problem requires to find a point x* and the value f(x*) being the global (i.e., the deepest) minimum of a function f(x) over an N-dimensional domain D, where f(x) can be non-differentiable, multiextremal, hard to evaluate even at one point (evaluations of f(x) are expensive), and given as a “black box”. Therefore, traditional local optimization methods cannot be used in this situation.