The asymptotic error of chaos expansion approximations for stochastic differential equations (Q2326537): Difference between revisions

The authors consider a one dimensional continuous stochastic process \(X_t\) that satisfies the stochastic differential equation \[ dX_t = b(t,X_t)dt + \sigma(t,X_t) dW_t, \quad X_0 = x_0, \] where \(W_t\) is a Brownian motion defined on a filtered probability space \((\Omega,\mathcal{F}, (\mathcal{F}_t), P)\). Let \(e_i\) be an orthonormal basis of the separable Hilbert space \(L^2(0,T])\). If \(X_t \in L^2(\Omega,\mathcal{F}, P)\), then it admits the chaotic expansion \[ X_t = \sum_{a \in\mathcal{I}} x_a(t) \Psi^a, \] where \(x_a\) are deterministic functions, \(\Psi^a\) are mutually orthogonal random projections that are associated to the basis \(e_i\) and the index \(\mathcal{I}\) is defined via \[ \mathcal{I} = \{ a = a_i : a_i \mathbb{N}_0 \text{ and almost all } a_i's \text{ are } 0 \}. \] They consider the approximation \[ X_t^{p,k} = \sum_{a \in\mathcal{I}_{p,k}} x_a(t) \Psi^a, \] where the subset \(\mathcal{I}_{p,k} \subseteq\mathcal{I}\) refers to using orthogonal projections \(\Psi^a\) only with respect to the first \(k\) basis elements \(e_i\) and only up to the \(p\)th order Wiener chaos.