The moment problem for continuous positive semidefinite linear functionals (Q1936506): Difference between revisions
From MaRDI portal
Added link to MaRDI item. |
Changed an Item |
||
Property / describes a project that uses | |||
Property / describes a project that uses: GloptiPoly / rank | |||
Normal rank |
Revision as of 07:54, 29 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The moment problem for continuous positive semidefinite linear functionals |
scientific article |
Statements
The moment problem for continuous positive semidefinite linear functionals (English)
0 references
5 February 2013
0 references
Let \(\mathbb{R}[\underline{X}]=\mathbb{R}[X_1,\ldots,X_n]\) be the real polynomial ring in \(n\) variables. Given a closed set \(K\subset \mathbb{R}^n\), the \(K\)-moment problem is the question whether a linear functional \(\ell :\mathbb{R}[\underline{X}]\to\mathbb{R}\) is representable as integration with respect to a positive Borel measure on \(K\). A necessary condition is obviously that \(\ell(f)\geq 0\) for all \(f\in\mathrm{Psd}(K)=\{g\in \mathbb{R}[\underline{X}]: g\geq 0\mathrm{ on }K\}\). By Haviland this condition is also necessary. However, this condition is in general impractical since the cone \(\mathrm{Psd}(K)\) is in general not finitely generated. But often this problem can be reduced to a finitely generated quadratic module of \(\mathbb{R}[\underline{X}]\): Assume that \(K\) is a compact basic closed semialgebraic set and that \(S=\{g_1,\ldots,g_s\}\) is a representation of \(K\) (i.e. \(K=\{x\in\mathbb{R}^n: g_1(x)\geq 0,\ldots,g_s(x)\geq 0\}\)) containing \(N-\sum{x_i^2}\) for some \(N\in\mathbb{N}\) (expressing that \(K=K_S\) is bounded). Then by Putinar's Archimedean Positivstellensatz it is enough to check the nonnegativity of the linear functional \(\ell\) on the finitely generated quadratic module \(M_S=\{\sigma_1g_1+\ldots+\sigma_sg_s: \sigma_1,\ldots,\sigma_s\mathrm{ sums of squares}\}\). This follows from the following topological consequence of Putinar's result: \(\mathrm{Psd}(K_S)\subseteq\overline{M_S}^\varphi\) where \(\varphi\) denotes the finest locally convex topology on \(\mathbb{R}[\underline{X}]\). In the literature also coarser topologies as the \(l_1\)-topology on \(\mathbb{R}[\underline{X}]\) are considered to study moment problems for example for \(l_1\)-continuous linear functionals. In the paper under review, the authors generalize this setting to arbitrary locally convex topologies. It is observed that given a locally convex topology \(\tau\), a closed subset \(K\) of \(\mathbb{R}^n\) and a cone \(C\) in \(\mathbb{R}[\underline{X}]\), any \(\tau\)-continuous linear functional that is nonnegative on \(C\) is integration with respect to a positive Borel measure on \(K\) if \(\mathrm{Psd}(K)\subseteq \overline{C}^\tau\). The closure of the cone of sums of squares \(\sum\mathbb{R}[\underline{X}]^2\) (i.e. of the smallest quadratic modul) is computed for the weighted \(p\)-norm-topologies and the topology of coefficientwise convergence.
0 references
positive polynomials
0 references
sums of squares
0 references
real algebraic geometry
0 references
moment problem
0 references
weighted norm topologies
0 references