Publication
In the classical NP-hard (metric) π-median problem, we are given a set of π clients and centers with metric distances between them, along with an integer parameter π β₯ 1. The objective is to select a subset of π open centers that minimizes the total distance from each client to its closest open center. In their seminal work, Jain, Mahdian, Markakis, Saberi, and Vazirani presented the Greedy algorithm for facility location, which implies a 2-approximation algorithm for π-median that opens π centers in expectation. Since then, substantial research has aimed at narrowing the gap between their algorithm and the best achievable approximation by an algorithm guaranteed to open exactly π centers, as required in the π-median problem. During the last decade, all improvements have been achieved by leveraging their algorithm (or a small improvement thereof), followed by a second step called bi-point rounding, which inherently adds an additional factor to the approximation guarantee. Our main result closes this gap: for any π > 0, we present a (2 +π)-approximation algorithm for the π-median problem, improving the previous best known approximation factor of 2.613. Our approach builds on a combination of two key algorithms. First, we present a non-trivial modification of the Greedy algorithm that operates with only π(logπ/π2) adaptive phases. Through a novel walkbetween-solutions approach, this enables us to construct a (2 + π)-approximation algorithm for π-median that consistently opens at most π +π(logπ/π2) centers: via known results, this already implies a (2 + π)-approximation algorithm that runs in quasi-polynomial time. Second, we develop a novel (2 + π)-approximation algorithm tailored for stable instances, where removing any center from an optimal solution increases the cost by at least an Ξ©(π3/logπ) fraction. Achieving this involves several ideas, including a sampling approach inspired by the π-means++ algorithm and a reduction to submodular optimization subject to a partition matroid. This allows us to convert the previous result into a polynomial time algorithm that opens exactly π centers while maintaining the (2+π)-approximation guarantee.