Section: Highlights of the Year

Highlights of the Year

Fast entropic methods for optimal transport problems: In a series of papers [19] [34] [10] [15] , MOKAPLAN's team members derived a new class of algorithm to obtain efficient approximations of the solution to various problems related to OT (including barycenters, Euler equation, unbalanced problems, gradient flows). This method makes use of entropic regularization and first order optimization method for the Kullback-Leibler divergence. See Section  6.3 for details about the software output.

Relaxing the mass conservation constraints: Our team derived a new theoretical and numerical framework to deal with “unbalanced” optimal transport problems [38] , [39] . This contribution is a breakthrough that will open the door to application in image processing and machine learning. See Section  7.6 for more details.