Finite dimensional approximation of diffusion processes on infinite dimensional spaces

From MaRDI portal





Let \(H\), \(E\) be separable Hilbert spaces, such that \(H\) is densely, continuously embedded in \(E\), through a Hilbert-Schmidt operator. A Dirichlet form on a space \(L^{2}(E, \mu )\) is defined as the closure of the symmetric, positive definite bilinear form \[ {\mathcal E} (u, v) := \int _{E} \langle A(z)\nabla u(z), \nabla v(z) \rangle_{H} \mu (dz) \] (where \(A(z)\) is a self adjoint and positive definite operator on \(H\)). Using projections from \(E\) onto finite-dimensional subspaces of \(H\), for each \(n\) a Dirichlet form \({\mathcal E}^{(n)}\) on \(L^{2}({\mathbb{R}}^n;\mu_n)\) is constructed. In turn, it gives rise to a probability measure \(P_{n}\) on the space of sample paths on \(E\) (the law of a finite-dimensional diffusion). The main theorem asserts that the sequence \((P_{n})\) converges weakly to a probability measure \(P\). Moreover, there exists a diffusion process \({\mathbf M} = (\Omega , {\mathcal F}, (X_{t})_{t \geq 0}, (P_{z})_{z\in E})\) on \(E\) , associated with the initial Dirichlet form \({\mathcal E}\), such that \(P = \int _{E} P_{z} \mu (dz)\) (the existence of \({\mathbf M}\) is known in a more general setting). As a by-product, the Hölder continuity of the sample paths is obtained. Finally, the paper discusses some important examples of infinite-dimensional diffusions, to which the above results apply.




Cited in
(22)








This page was built for publication: Finite dimensional approximation of diffusion processes on infinite dimensional spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4363320)