EPTAS for stable allocations in matching games (2107.07440v2)

Abstract: Gale-Shapley introduced a matching problem between two sets of agents where each agent on one side has a preference over the agents of the other side and proved algorithmically the existence of a pairwise stable matching (i.e. no uncoupled pair can be better off by matching). Shapley-Shubik, Demange-Gale, and many others extended the model by allowing monetary transfers. In this paper, we study an extension where matched couples obtain their payoffs as the outcome of a strategic game and more particularly a solution concept that combines Gale-Shapley pairwise stability with a constrained Nash equilibrium notion (no player can increase its payoff by playing a different strategy without violating the participation constraint of the partner). Whenever all couples play zero-sum matrix games, strictly competitive bi-matrix games, or infinitely repeated bi-matrix games, we can prove that a modification of some algorithms in the literature converge to an $\varepsilon$-stable allocation in at most $O(\frac{1}{\varepsilon})$ steps where each step is polynomial (linear with respect to the number of players and polynomial of degree at most 5 with respect to the number of pure actions per player).