Measurable One-to-One Selections and Transition Kernels (Q3691985)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3914637
Language Label Description Also known as
default for all languages
No label defined
    English
    Measurable One-to-One Selections and Transition Kernels
    scientific article; zbMATH DE number 3914637

      Statements

      Measurable One-to-One Selections and Transition Kernels (English)
      0 references
      0 references
      0 references
      1985
      0 references
      analytic spaces
      0 references
      probability
      0 references
      Borel one-to-one selection
      0 references
      Borel isomorphism
      0 references
      Martin's axiom
      0 references
      universally measurable one-to-one selection
      0 references
      transition kernels
      0 references
      The main result of the paper reads as follows: Let X and Y be analytic spaces, \(R\subset X\times Y\) a Borel set, and \(\mu\) a probability measure on Y. If \(R^ y=\{x\in X:(x,y)\in R\}\) is uncountable for \(\mu\)-a.e. \(y\in Y\) then there exists a Borel set \(B\subset Y\) with \(\mu (B)=1\) and a Borel one-to-one selection f:\(B\to X\) for R. If, in addition, X carries a probability measure \(\lambda\) such that, for \(\lambda\)-a.e. \(x\in X\), the set \(R_ x=\{y\in Y:(x,y)\in R\}\) is uncountable then there are Borel sets \(B\subset Y\) and \(A\subset X\) of full \(\mu\)-resp. \(\lambda\)-measure and a Borel isomorphism from A onto B whose graph is contained in R. Moreover it is shown that there need not be an everywhere defined Borel one-to-one selection even if \(X=Y=[0,1]\), \(\mu\) equals Lebesgue measure, and \(\mu (R^ y)=1\) for all \(y\in Y\). In contrast to this result it is proved that, assuming Martin's axiom, there is an everywhere defined universally measurable one-to-one selection for R provided \(\lambda (R^ y)>0\) for all \(y\in Y\) where \(\lambda\) is an atomless probability measure on X. The methods of proof rely on the following fact concerning Borel measurable many-to-one selections which is proved using transition kernels: If R is uncountable for \(\mu\)-a.e. \(y\in Y\), \(\lambda\) is any probability measure on X then there exists a \(\lambda\)-nullset \(A\subset X\) and a Borel selection g:\(A\to Y\) for R with \(g^{-1}(y)\) uncountable for \(\mu\)-a.e. \(y\in Y\).
      0 references

      Identifiers

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