CRM: Centro De Giorgi

This is the old version of the CRM site. Please use the new site on the page crmdegiorgi.sns.it

logo sns
Optimal Transportation and Applications

Discretizing Wasserstein gradient flows: simple examples with surprising properties

speaker: Daniel Matthes (Technische Universität München)

abstract: Four evolution equations are considered: the linear heat equation, the quadratic porous medium equation, the fourth order DLSS equation and a non-local QDD model. These PDEs have in common that in addition to their gradient flow structure in L2-Wasserstein, they have further substantial analytical properties that facilitate the analysis, like contractivity in the Hellinger metric or in a novel variant of the martingale OT distance. For each of the PDEs, we propose a spatial discretization that is simple but provides approximative solutions which are global, positive, and preserve an (at least in the case of DLSS: surprisingly large) number of the aforementioned structural properties. The latter are key e.g. in the convergence analysis, or for proving long-time asymptotics directly on the discrete level. This is joint work with Giuseppe Savare, Andre Schlichting, and Eva-Maria Rott.


timetable:
Tue 3 Dec, 12:00 - 12:45, Aula Dini
<< Go back