Metric entropy and the small ball problem for Gaussian measures (Q1316388): Difference between revisions

Let \(\mu\) be a centered Gaussian measure on a real separable Banach space \(B\) with norm \(\| \cdot \|\) and dual \(B^*\), \(S : B^* \to B\) be the covariance operator of \(\mu\). Then it is known that \[ \lim_{t \to \infty} t^{-2} \log \mu \biggl( \bigl \{x : \| x \| \geq t \bigr \} \biggr) = - (2 \| S \|)^{-1}. \] More delicate is the study of the behaviour of the function \[ \varepsilon \to \varphi (\varepsilon) = - \log \mu \biggl( \bigl \{x : \| x \| \leq \varepsilon \bigr\} \biggr) \] when \(\varepsilon \to 0\). Several results in this direction are known, f.e. when \(B\) is a Hilbert space and \(S\) has a special sequence of eigenvalues or when \(B=C[0,1]\) and \(\mu\) is Wiener measure. With the operator \(S\) one can associate in a canonical way a compact subset \(K \subset B\). The author's study of the behavior of \(\varphi (\varepsilon)\) for small \(\varepsilon\) in terms of metric entropy \(H (\varepsilon,K)\) of \(K\).