Abstract
Consider a matching problem, in which agents arrive to a marketplace over time and leave after some time periods. Agents can only be matched while present in the marketplace. Each pair of agents can yield a different match value, and a social planner seeks to maximize the total value from matches over a finite time horizon. First we study the case in which vertices arrive in an adversarial order. For the case when agents depart in the order of arrival, we provide a randomized [Formula: see text]-competitive algorithm. When departure times are drawn independently from a distribution with nondecreasing hazard rate, we establish a [Formula: see text]-competitive algorithm. When the arrival order is chosen uniformly at random and agents leave after a fixed number of time periods, a batching algorithm, which computes a maximum-weighted matching periodically, is shown to be 0.279-competitive.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have