MOKAPLAN is a joint INRIA-CNRS-Université Paris-Dauphine INRIA Team.
We develop numerical methods, algorithms and softwares along four axes:
- Variational problems, in particular those related to optimal transport (MoKa stands for Monge-Kantorovitch);
- Application of optimal transport numerics to non-variational and non-convex problems (market design, equilibrium and transport);
- Inverse problems with structured priors (off-the-grid sparse recovery, Burer-Monteiro approaches ) for imaging sciences;
- Geometric variational problems (approximation of measures with geometric constraints, discretization of singular measures).
A common denominator between all these axes is the need to design numerical methods for working with measures.
The range of potential outputs is large and we focus on selected applications in natural and social sciences.