1: \begin{abstract}

2:   Many problems in geometric optics or convex geometry can be recast

3:   as optimal transport problems and a popular way to solve these

4:   problems numerically is to assume that the source probability

5:   measure is absolutely continuous while the target measure is

6:   finitely supported. We introduce a damped Newton's algorithm for

7:   this type of problems, which is experimentally efficient, and we

8:   establish its global linear convergence for cost functions

9:   satisfying an assumption that appears in the regularity theory for

10:   optimal transport.

11: \end{abstract}

12: