scientific article; zbMATH DE number 1188839

From MaRDI portal
Publication:3838522





In this paper the authors analyze the reconstructed solution based on noisy Hermite-Fourier coefficients of an unknown function of a weighted Sobolev space \(H^p_w\), with the weight \(w(x)=\exp(-x^2/2)\). More specific, the Hermite-Fourier coefficients \(f_k=\langle f,H_k\rangle_w\) of a function \(f\in H^p_w\) (where \(H_k\) are the orthonormalized Hermite polynomials) are affected additively by the noise \(\epsilon_k\): \(f^{\epsilon}_k = f_k+\epsilon_k\). To reconstruct the signal \(f\), one uses a linear estimator based on the measurements \(f^{\epsilon}_k\) by solving the minimization problem \[ \min_c \parallel f^{\epsilon}-c\parallel_{l^2}^2 + \lambda \parallel \sum_{k}c_k H_k\parallel^2_{H^p_w}. \] For an optimal choice of \(\lambda\), the generalized cross validation criterion is shown to be asymptotically optimal and the reconstruction error is asymptotically bounded (whereas for \(c=f^{\epsilon}\) it is unbounded). For the entire collection see [Zbl 0892.00039].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3838522)