Restricted Max-Min Allocation: Integrality Gap and Approximation Algorithm

Abstract

Given a set of players P, a set of indivisible resources R, and a set of non-negative values $\{v_{pr}\}_{p\in P, r\in R}$, an allocation is a partition of R into disjoint subsets $\{C_p\}_{p \in P}$ so that each player p is assigned the resources in $C_p$. The max-min fair allocation problem is to determine the allocation that maximizes $\min _p \sum _{r\in C_p}v_{pr}$. In the restricted case of this problem, each resource r has an intrinsic value $v_r$, and $v_{pr} = v_r$ for every player p who desires r and $v_{pr} = 0$ for every player p who does not. We study the restricted max-min fair allocation problem in this paper. For this problem, the configuration LP has played an important role in estimating and approximating the optimal solution. Our first result is an upper bound of $3\frac{21}{26}$ on the integrality gap, which is currently the best. It is obtained by a tighter analysis of the local search of Asadpour et al. [TALG’12]. It remains unknown whether this local search runs in polynomial time or not. Our second result is a polynomial-time algorithm that achieves an approximation ratio of $4 + \delta$ for any constant $\delta \in (0,1)$. Our algorithm can be seen as a generalization of the aforementioned local search.