Improved Approximation and Scalability for Fair Max-Min Diversification
Abstract
Given an -point metric space where each point belongs to one of different categories or groups and a set of integers , the fair Max-Min diversification problem is to select points belonging to category , such that the minimum pairwise distance between selected points is maximized. The problem was introduced by Moumoulidou et al. [ICDT 2021] and is motivated by the need to down-sample large data sets in various applications so that the derived sample achieves a balance over diversity, i.e., the minimum distance between a pair of selected points, and fairness, i.e., ensuring enough points of each category are included. We prove the following results: 1. We first consider general metric spaces. We present a randomized polynomial time algorithm that returns a factor -approximation to the diversity but only satisfies the fairness constraints in expectation. Building upon this result, we present a -approximation that is guaranteed to satisfy the fairness constraints up to a factor for any constant . We also present a linear time algorithm returning an approximation with exact fairness. The best previous result was a approximation. 2. We then focus on Euclidean metrics. We first show that the problem can be solved exactly in one dimension. For constant dimensions, categories and any constant , we present a approximation algorithm that runs in time where . We can improve the running time to at the expense of only picking points from category . Finally, we present algorithms suitable to processing massive data sets including single-pass data stream algorithms and composable coresets for the distributed processing.
Cite
@article{arxiv.2201.06678,
title = {Improved Approximation and Scalability for Fair Max-Min Diversification},
author = {Raghavendra Addanki and Andrew McGregor and Alexandra Meliou and Zafeiria Moumoulidou},
journal= {arXiv preprint arXiv:2201.06678},
year = {2022}
}
Comments
To appear in ICDT 2022