MN - Modèles Numériques

Interacting Systems and Optimal Transportation, Applications to Computational Economics. – ISOTACE

Submission summary

The use of partial differential equations (PDE) in Economics and more generally social sciences is a relatively young field of research.
Optimal Transportation* (OT) and Mean field Games (MFG) theories have, in this context, recently received a lot of attention. The
"Computational Fluid Dynamic" (CFD) formulation of OT is intimately linked to the more general PDE structure of MFG.
Both OT and MFG mathematically underlie our proposal.

The corresponding research effort to produce numerical analysis and algorithms remains scattered and largely academic.
Meanwhile, the number of modeling issues and associated mathematical and numerical analysis questions is expanding at a rapid pace.
The design of efficient solvers is a bottleneck that can be addressed only by numerical analysts and specialists of scientific computing together
with economists and mathematicians; without such experimental and numerical development, it will be difficult for these numerical models to become industrial simulation tools. Numerical simulations are also a key ingredient in proposing and studying new academic models.

The proposed consortium brings together economists, applied mathematicians and specialists of scientific computing in an attempt to bridge
the gap between theory and numerics, academic and industrial applications.
Producing reusable code is a common approach for numerical methods targeting "hard" science applications. It is known that making public domain implementation available stimulates research and helps technology transfer to the industry.
We therefore plan to deliver two libraries of solvers dedicated respectively to MFG/Dynamic OT and Static OT.

MFG and OT based models for crowd motion under congestion are mature enough to propose a multi-solver/multi-model software development task.
It will be a "real wold" test for the integration of solver libraries mentioned above.
The software objectives are both academic to understand/compare these modern congestion models and also to be used as a prototype for industrial collaborations**.

With a different perspective the academic environment of the consortium will help/support
the development by our industrial partner ZELIADE of a software with application to model-free bounds in derivative pricing.
It is a completely new application of OT and is expected to have a direct industrial impact in finance.

In more academically oriented approach, we finally propose to explore a set of new models in economy where the MFG and OT
libraries will provide useful base for numerical simulations, in particular :
- Many relevant markets are markets of indivisible goods characterized by a certain quality: houses, jobs, marriages... On the theoretical side,
finding equilibria in such markets is equivalent to solving a certain optimal transport problem (where the cost function depends on the sellers
and buyers preferences). On the empirical side, this allows for trying to recover information on the preferences from observed matching; this is an
inverse problem.
- The central principal agent model addresses the problem of a principal minimizing cost under the incentive constraint faced by an agent with multivariate type in the case of a simple bilinear utility. The problem can be reduced to the maximization of an integral functional subject to a convexity constraint. Convexity/well-posedness for more general utilities has been recently obtained using OT techniques.

* Also known as the L2-Monge Kantorovich problem.
** Reaching out to public/industrial related effort is an explicit component of the coordination task.
Our industrial partner MFGlabs will also be active on the valorization of this task.

Project coordination

jean-david BENAMOU (Institut National de Recherche en Informatique et Automatique) –

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.


MOKAPLAN Institut National de Recherche en Informatique et Automatique
LMO Universite Paris Sud
CMAP Ecole Polytechnique
LJLL-P7 Laboratoire Jacques-Louis Lions

Help of the ANR 620,861 euros
Beginning and duration of the scientific project: October 2012 - 48 Months

Useful links

Explorez notre base de projets financés



ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter