Popularity, Mixed Matchings, and Self-Duality

Abstract

Our input instance is a bipartite graph G where each vertex has a preference list ranking its neighbors in a strict order of preference. A matching M is popular if there is no matching N such that the number of vertices that prefer N to M outnumber those that prefer M to N. Each edge is associated with a utility and we consider the problem of matching vertices in a popular and utility-optimal manner. It is known that it is NP-hard to compute a max-utility popular matching. So we consider mixed matchings: a mixed matching is a probability distribution or a lottery over matchings. Our main result is that the popular fractional matching polytope PG is half-integral and in the special case where a stable matching in G is a perfect matching, this polytope is integral. This implies that there is always a max-utility popular mixed matching which is the average of two integral matchings. So in order to implement a max-utility popular mixed matching in G, we need just a single random bit. We analyze the popular fractional matching polytope whose description may have exponentially many constraints via an extended formulation with a linear number of constraints. The linear program that gives rise to this formulation has an unusual property: self-duality. The self-duality of this LP plays a crucial role in our proof. Our result implies that a max-utility popular half-integral matching in G and also in the roommates problem (where the input graph need not be bipartite) can be computed in polynomial time.