Perturbed sums-of-squares theorem for polynomial optimization and its applications

Masakazu Muramatsu, Hayato Waki, Levent Tunçel

    Research output: Contribution to journalArticle

    1 Citation (Scopus)

    Abstract

    We consider a property of positive polynomials on a compact set with a small perturbation. When applied to a polynomial optimization problem (POP), the property implies that the optimal value of the corresponding SemiDefinite Programming (SDP) relaxation with sufficiently large relaxation order is bounded from below by (f∗-ε) and from above by f∗+ ε(n + 1), where f∗is the optimal value of the POP. We propose new SDP relaxations for POP based on modifications of existing sums-of-squares representation theorems. An advantage of our SDP relaxations is that in many cases they are of considerably smaller dimension than those originally proposed by Lasserre. We present some applications and the results of our computational experiments.

    Original languageEnglish
    Pages (from-to)134-156
    Number of pages23
    JournalOptimization Methods and Software
    Volume31
    Issue number1
    DOIs
    Publication statusPublished - Jan 2 2016

      Fingerprint

    All Science Journal Classification (ASJC) codes

    • Software
    • Control and Optimization
    • Applied Mathematics

    Cite this