Semi-discrete multi-to -one dimensional variational problems

TL;DR

This paper introduces a new approach for semi-discrete variational problems in economic matching and game theory, providing simplified solutions, efficient algorithms, and convergence guarantees, with demonstrated superior performance in numerical experiments.

Contribution

It develops a novel theoretical framework for semi-discrete variational problems, leading to scalable algorithms with proven convergence, applicable to economic matching and related fields.

Findings

Solutions have a simple structure under certain conditions.

Algorithms outperform standard optimal transport solvers in large outcome spaces.

Numerical examples demonstrate significant efficiency gains.

Abstract

We study a class of semi-discrete variational problems that arise in economic matching and game theory, where agents with continuous attributes are matched to a finite set of outcomes with a one dimensional structure. Such problems appear in applications including Cournot-Nash equilibria, and hedonic pricing, and can be formulated as problems involving optimal transport between spaces of unequal dimensions. In our discrete strategy space setting, we establish analogues of results developed for a continuum of strategies in \cite{nenna2020variational}, ensuring solutions have a particularly simple structure under certain conditions. This has important numerical consequences, as it is natural to discretize when numerically computing solutions. We leverage our results to develop efficient algorithms for these problems which scale significantly better than standard optimal transport solvers, particularly when the number of discrete outcomes is large, provided our conditions are satisfied. We also establish rigorous convergence guarantees for these algorithms. We illustrate the advantages of our approach by solving a range of numerical examples; in many of these our new solvers outperform alternatives by a considerable margin.