Non-convex regularization of bilinear and quadratic inverse problems by tensorial lifting
For researchers in inverse problems and regularization theory, this work provides a unified framework that extends convex regularization to a broader class of non-linear problems, though the theoretical results are incremental as they generalize existing concepts.
The paper introduces the class of dilinear mappings to extend linear regularization theory to bilinear and quadratic inverse problems, including blind deconvolution and phase retrieval. By lifting these problems to linear ones on tensor spaces and using diconvex regularization, they establish convergence rates similar to the linear setting and validate them numerically on the deautoconvolution problem.
Considering the question: how non-linear may a non-linear operator be in order to extend the linear regularization theory, we introduce the class of dilinear mappings, which covers linear, bilinear, and quadratic operators between Banach spaces. The corresponding dilinear inverse problems cover blind deconvolution, deautoconvolution, parallel imaging in MRI, and the phase retrieval problem. Based on the universal property of the tensor product, the central idea is here to lift the non-linear mappings to linear representatives on a suitable topological tensor space. At the same time, we extend the class of usually convex regularization functionals to the class of diconvex functionals, which are likewise defined by a tensorial lifting. Generalizing the concepts of subgradients and Bregman distances from convex analysis to the new framework, we analyse the novel class of dilinear inverse problems and establish convergence rates under similar conditions than in the linear setting. Considering the deautoconvolution problem as specific application, we derive satisfiable source conditions and validate the theoretical convergence rates numerically.