The best low rank approximation of a given matrix can be found through singular value decomposition. However due to the non-convexity of the problem incorporation of any additional priors is difficult. In this talk we show how to compute convex envelopes of a class of objective functions useful for low rank approximation. We derive efficient algorithms for evaluating the envelope and its proximal operator which enables large scale optimization in general convex frameworks.

This presentation is part of Minisymposium “MS37 - Sparse-based techniques in variational image processing (2 parts)”
organized by: Serena Morigi (Dept. Mathematics, University of Bologna) , Ivan Selesnick (New York University) , Alessandro Lanza (Dept. Mathematics, University of Bologna) .

Authors:

Carl Olsson (Chalmers University of Technology)

Keywords: