Paul Catala - A low-rank solver for off-the-grid Wasserstein group-Lasso
12 May 2020 –
Tuesday 12 May, 11h(FR)=12h(GR) at https://meet.jit.si/Aromath
Paul Catala (DMA ENS Paris)
A low-rank solver for off-the-grid Wasserstein group-Lasso We consider the problem of simultaneously recovering pointwise sources across several similar tasks given some low-pass measurements. The group Lasso regularizes this problem by enforcing a common sparse support to the solutions in each task, which is often too restrictive in real applications. In this talk, I will introduce a new off-the-grid (i.e. without spatial discretization) solver for this multi-task recovery, which integrates a Wasserstein distance between the recovered sources. This solver uses a semidefinite programming relaxation based on Lasserre's hierarchy. Extension to the recovery of non-sparse objects will be discussed.