Non-asymptotic convergence bounds for Wasserstein approximation using point clouds

Our main result shows that, under the two above mentioned assumptions (that is, ρ is close to a constant in C 2 and the initial datum is smooth and increasing) the discrete and

Minibatch Wasserstein While the entropic loss has better computational complexity than the orig- inal Wasserstein distance, it is still challenging to compute it for a large

To deal with such degeneracies, we first give an abstract existence and uniqueness result for viscosity solutions of non-local degenerate Hamiltonians, satisfying suitable

Abstract: This paper establishes a link between some space discretization strate- gies of the Finite Volume type for the Fokker-Planck equation in general meshes (Voronoï

The large time behaviour of nonnegative solutions to a quasilinear degenerate diffusion equation with a source term depending solely on the gradient is investigated.. After a

It is a stochastic variance- reduced proximal-gradient type algorithm built on Stochastic Path Integral Differential EstimatoR (SPIDER), an algorithm known to achieve

Indeed, since the pioneering works [12] on the linear Fokker-Planck equation and [15, 16] on the porous medium equation, several equations have been interpreted as gradient flows

Moreover, under some regularity assumption on the initial datum u 0 , such a variational problem allows to define a weak formulation of the Hele-Shaw flow [9, 11] (see also [12] for