Approximation complexity of tensor product-type random fields with heavy spectrum (Q2439845): Difference between revisions

A sequence \(X_d(t)\) of Gaussian tensor product-type random fields is considered, where \(X_d(t)\) contains independent standard Gaussian random variables and all eigenvalues and eigenfunctions of the covariance operator of the process \(X_1\). Here, \(t\in [0,1]^d\) and \(d\in\mathbb{N}\). For each \(d\in\mathbb{N}\), the sample paths of \(X_d\) almost surely belong to \(L_2([0,1]^d)\) with norm \(\|.\|_{2,d}\). Furthermore, the so-called eigenpairs of the covariance operator of \(X_d\) are considered. Finally, the random fields \(X_d\) are approximated by the finite sums \(X^{(n)}_d\) corresponding to the \(n\) maximal eigenvalues of the covariance operator of \(X_d\). The main results contain investigations on the logarithmic asymptotics of the average approximation complexity and of the probabilistic approximation complexity as the parametric dimension \(d\to\infty\).