Phd Thesis
Inserted: 2 aug 2019
Last Updated: 2 aug 2019
Year: 2019
Abstract:
The Wasserstein space is the space of probability measures over a given domain endowed with the quadratic Wasserstein distance. In this work, we study variational problems where the unknowns are mappings valued in the Wasserstein space.
When the source space is a segment, i.e. when the unknowns are curves valued in the Wasserstein space, we are interested in models where, in addition to the action of the curves, there are some terms which penalize congested configurations. We develop techniques to extract regularity from the minimizers thanks to the interplay between optimal density evolution (minimization of the action) and penalization of congestion, and we apply them to the study of Mean Field Games and the variational formulation of the Euler equations.
When the source space is no longer a segment but a domain of a Euclidean space, we consider only the Dirichlet problem, i.e. the minimization of the action (which can be called the Dirichlet energy) among mappings sharing a fixed value on the boundary of the source space. The solutions are called harmonic mappings valued in the Wasserstein space. We prove that the different definitions of the Dirichlet energy in the literature turn out to be equivalent; that the Dirichlet problem is well-posed under mild assumptions; that the superposition principle fails if the source space is no longer a segment; that a sort of maximum principle holds; and we provide a numerical method to compute these harmonic mappings.
Download: