Approximating Positive Polynomials Using Sums of Squares

From MaRDI portal





Each finite subset \(S\) of a finitely generated algebra \(A= \mathbb{R}[x_1,\dots, x_n]\) defines the set of nonnegativity \(X_S= \{t\in\mathbb{R}^n\mid\forall s\in S: s(t)\geq 0\}\), the quadratic module \(M_S\) (which consists of the sums \(\sum_{s\in S\cup\{1\}} \sigma_s\cdot s\) where the \(\sigma_s\) are sums of squares in \(A\)), the preordering \(T_S\) (which is the quadratic module defined by the set of square free products of elements of \(S\)), and the cone \(\overline{T_S}\) of polynomial functions that are positive semi-definite on \(X_S\). The inclusions \(M_S\subseteq T_S\subseteq\overline{T_S}\) hold trivially. It is a basic task of real algebra to decide whether a given polynomial function belongs to one of these sets. The question is closely related with the moment problem from functional analysis. The present paper studies the relationship between \(M_S\) and \(\overline{T_S}\). The following is a typical result: If \(p\in M_S\) and if \(1+ p\) is a unit in \(A\) then the following statements are equivalent for an element \(a\in A\): \(\bullet\) \(a\in\overline{T_S}\), and \(\bullet\) \(\exists k\in\mathbb{N}\forall 0< \varepsilon\in \mathbb{Q}: a+ \varepsilon\cdot(1+ p)^k\in M_S\).











This page was built for publication: Approximating Positive Polynomials Using Sums of Squares

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4465076)