The hierarchy of local minimums in polynomial optimization (Q2349136)

The author studies how to compute the hierarchy of critical values and local minimums of a polynomial, by using semidefinite relaxations. For this purpose, \(H\)-mimimizers (points satisfying the first- and second-order necessary optimality conditions) are computed. To compute each \(H\)-minimum, a sequence of semidefinite relaxations, based on optimality conditions is constructed. It is proved that each sequence has finite convergence to an \(H\)-minimum, under some generic conditions. A procedure for computing all \(H\)-minimums is presented. When there are equality constraints, similar results for computing the hierarchy of critical values and local minimums are obtained.