SOS approximations of nonnegative polynomials via simple high degree perturbations (Q883108)

From MaRDI portal





scientific article
Language Label Description Also known as
default for all languages
No label defined
    English
    SOS approximations of nonnegative polynomials via simple high degree perturbations
    scientific article

      Statements

      SOS approximations of nonnegative polynomials via simple high degree perturbations (English)
      0 references
      0 references
      0 references
      31 May 2007
      0 references
      Define polynomials \(\Theta_r=1+\sum_{j=1}^n X_j^{2r} \in \mathbb R[X]=\mathbb R[X_1,\dots,X_n],\) \(r=0,1,2,\ldots\). One of the main results (C3.3+T3.9) is Theorem: Given any \(\epsilon >0\) and a polynomial \(f\in \mathbb R[X]\) such that \(f| [-1,1]^n \geq 0.\) Then for \(r\) large enough the polynomial \(f_{\epsilon,r}= f+\epsilon \Theta_r\) is s.o.s. Integer \(r\) depends only on \(\epsilon,\) the dimension \(n,\) and the size of the coefficients of \(f.\) This provides a (in the worst cases not constructive) certificate of nonnegativity of \(f\) on \([-1,1]^n;\) it also says that every polynomial nonnegative on a (0-centered) \(n\)-cube can be approximated uniformly by sums of squares. This complements results of \textit{J. B. Lasserre} [SIAM J. Optim. 16, No. 3, 751--765 (2006; Zbl 1129.12003)] where similar certificates for polynomials nonnegative on \(\mathbb R^n\) are proven, and results of Berg, Christensen, and Ressel [see \textit{C. Berg}, Proc. Symp. Appl. Math. 37, 110--124 (1987; Zbl 0636.44007)], where denseness of s.o.s. approximations is shown, but no explicit approximation sequences are given. An analogous result is shown for nonnegativity on certain basic semialgebraic sets: for a finite set \(S\) of polynomials, having the strong moment property, let \(K_S\) be the associated basic closed semialgebraic set, and \(T_S\) the preorder defined by \(S.\) Suppose \(K_S\) has nonempty interior. Then polynomials \(\theta_r= \sum_{i=1}^n\sum_{k=0}^r \frac{X_i^{2k}}{k!} \) (used in place of the \(\Theta_r\)) can serve to establish for polynomials \(f\) and large \(r\) that \(f_{\epsilon,r}\in T_S,\) and hence \(f| K_S\geq 0.\)
      0 references
      real algebraic geometry
      0 references
      positive polynomials
      0 references
      sum of squares
      0 references
      semidefinite programming
      0 references
      moment problem
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references