Dual estimates in multiextremal problems (Q1201906)

The paper proposes a technique of improving the dual estimates in nonconvex multiextremal problems of mathematical programming by adding some additional constraints which are the consequences of the original constraints as follows: Let a (bounded-from-below) polynomial \(P(x_ 1,x_ 2,\dots,x_ n)\) be given and let \(P^*\) be the value of the polynomial at the global minimum point. By introducing new variables and making use of quadratic substitutions of the form: \(x^ 2_ i=y_ i\); \(x_ j x_ k=z_{kj}\), and so forth, we can reduce the minimization problem for the polynomial \(P(x_ 1,x_ 2,\dots,x_ n)\) to a quadratic extremal problem with constraints in the form of equalities. This technique is used for problems of finding the global optimum of polynomial functions, and extremal quadratic and Boolean quadratic problems. Also, the paper considers an ecological multiextremal problem and gives an algorithm for finding the dual estimate for it. The algorithm is based on a scheme of decomposition and nonsmooth optimization methods.