Abstract:
We study measures on a real separable Hilbert space $E$ that are invariant under
translations by arbitrary vectors in $E$. We define the Hilbert space $\mathcal H$ of
complex-valued functions on $E$ square-integrable with respect to some translation-invariant measure $\lambda$. We determine the expectations of the operators of shift
by random vectors whose distributions are given by semigroups (with respect to
convolution) of Gaussian measures on $E$. We prove that these expectations form a semigroup of self-adjoint contractions on $\mathcal H$. We obtain a criterion for the
strong continuity of such semigroups and study the properties of their generators
(which are self-adjoint generalizations of Laplace operators to the case of functions
of infinite-dimensional arguments). We introduce analogues of Sobolev spaces and
spaces of smooth functions and obtain conditions for the embedding and dense embedding of
spaces of smooth functions in Sobolev spaces. We apply these function spaces
to problems of approximating semigroups by the expectations of random processes and
study properties of our generalizations of Laplace operators and their fractional
powers.
Keywords:translation-invariant measure on a Hilbert space, Laplace operator on an infinite-dimensional space, Sobolev spaces, embedding theorems, random walks.