104
4

Regularization with Approximated L2L^2 Maximum Entropy Method

Abstract

We tackle the inverse problem of reconstructing an unknown finite measure μ\mu from a noisy observation of a generalized moment of μ\mu defined as the integral of a continuous and bounded operator Φ\Phi with respect to μ\mu. When only a quadratic approximation Φm\Phi_m of the operator is known, we introduce the L2L^2 approximate maximum entropy solution as a minimizer of a convex functional subject to a sequence of convex constraints. Under several assumptions on the convex functional, the convergence of the approximate solution is established and rates of convergence are provided.

View on arXiv
Comments on this paper