Uniqueness in Bounded Moment Problems

From MaRDI portal





Let \((X,{\mathfrak A},\mu)\) be a \(\sigma\)-finite measure space and \({\mathcal K}\) be a linear subspace of \({\mathcal L}_ 1(\mu)\) such that supp \({\mathcal K}=X\). A set \(A\in{\mathfrak A}\) is called \({\mathcal K}\)-determined within the class \({\mathcal G}\) of measurable functions \(g\) bounded by \(0\leq g\leq 1\) if one can conclude from \(g\in{\mathcal G}\) and \(\int fgd\mu=\int f1_ Ad\mu\) for all \(f\in{\mathcal K}\) that \(g\) is the indicator function \(1_ A\) of \(A\). It is known that all sets \(A\) which admit a representation \(A=\{f\geq 0\}\) for some \(f\in{\mathcal K}\) are \({\mathcal K}\)-determined. But this condition is not necessary. The author now gives a complete characterization of \({\mathcal K}\)-determined sets \(A\): These are the sets such that \(A\) and its complement have a representation \(A=\{f^*>0\}\) and \(X\backslash A=\{f^*<0\}\) with some function \(f^*\) which is in a hull \({\mathcal K}^*\) of measurable extended real valued functions obtained as weak limits of functions in \({\mathcal K}\).











This page was built for publication: Uniqueness in Bounded Moment Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4039344)