Bias reduction in variational regularizationMS39

We present a two-step method to reduce bias in variational methods. After solving the standard variational problem, the idea is to perform a consecutive debiasing step minimizing the data fidelity on an appropriate model subspace. It is defined by Bregman distances using the subgradient appearing in the optimality condition of the variational method. This leads to a decomposition of the overall bias into two parts, model and method bias, of which we tackle the latter.

This presentation is part of Minisymposium “MS39 - Nonlinear Spectral Theory and Applications (part 2)
organized by: Aujol Jean-Francois (University of Bordeaux) , Gilboa Guy (Electrical Engineering Department, Technion) .

Camille Sutour (University Paris Descartes)