Regularization with Approximated Maximum Entropy Method

Abstract
We tackle the inverse problem of reconstructing an unknown finite measure from a noisy observation of a generalized moment of defined as the integral of a continuous and bounded operator with respect to . When only a quadratic approximation of the operator is known, we introduce the 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 arXivComments on this paper