0-1 laws for kernels of a linear uniform measure (Q1100577): Difference between revisions

The 0-1 laws for kernels of a linear uniform measure are investigated. The probability measure \(\mu\) on a locally convex Hausdorff space E is called uniform if for \(x_ n'\in E'\) (the dual) \(\mu (x;\quad x_ n'(x)\to 0)>0\) implies that \(x_ n'\to 0\) in probability [see \textit{R. M. Dudley} and \textit{M. Kanter}, Proc. Am. Math. Soc. 45, 245-252 (1974; Zbl 0297.60007) and the authors Hokkaido Math. J. 16, 213-222 (1987; Zbl 0637.28010)]. The kernel K(\(\mu)\) of \(\mu\) is defined by \[ K(\mu) = \cap \{Z;\quad \mu (Z)=1,\quad Z=\{x;\quad x_ n'(x)\to 0\},\quad x_ n'\in E'\}, \] for related notions see \textit{J. Hoffmann-Jørgensen} [Stud. Math. 61, 139-159 (1977; Zbl 0373.60014)], \textit{C. Borell} [Ark. Mat. 14, 79-92 (1976; Zbl 0331.60008)] and \textit{S. Chevet} [Lect. Notes Math. 860, 51-84 (1981; Zbl 0458.60005)]. The main result is as follows. If \(\mu\) is uniform, then \(\mu^*(K(\mu))=0\) or 1, where \(\mu^*\) is the outer measure. In the case of \(\mu^*(K(\mu))=1,\) K(\(\mu)\) is of finite dimension. The proof depends on the following Lemma: If \(\mu^*((E',\tau_{\mu})')=1,\) then \((E',\tau_{\mu})\) is locally convex nuclear and semi-metrizable, where \(\tau_{\mu}\) is the topology of convergence in measure [see \textit{S. Kwapien} and \textit{W. Smolenski}, Z. Wahrscheinlichkeitstheorie Verw. Geb. 59, 197-201 (1982; Zbl 0498.60004) and the authors, J. Multivariate Anal. 22, 65-73 (1987)].