The aim of this project is to implement various methods for the dynamic formulation of optimal transport, namely the optimization of the kinetic energy on the space of all solutions on the continuity equation. The most basic dynamic OT problem is the Benamou-Brenier formulation of the Wasserstein distance. Given probability densities
We will implement the algorithm to solve this problem and its generalizations.
For now, our main aim is to implement the dynamical formulation of the Wasserstein-Fisher-Rao optimal transport, where we solve the following problem:
We approach this problem in two ways: the proximal
algorithm, originally by Chizat et al. (2018), and the relaxed
algorithm.
Currently, we have finished implementing the relaxed
method for the unconstrained case.