On the Cost of Essentially Fair Clusterings

Bercea, Ioana O.; Groß, Martin; Khuller, Samir; Kumar, Aounon; Rösner, Clemens; Schmidt, Daniel R.; Schmidt, Melanie

doi:10.4230/lipics.approx-random.2019.18

Cited by 8 publications

(8 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second one is the practical Linear Program from [2] (Not the 3− approximation). Note that we did not include the algorithm from [6] as it TLE on almost all input instances as was the case for the 3−approximation from [2]. This is mainly due to them using Θ(N 2 ) LP variables.…”

Section: Methodsmentioning

confidence: 99%

“…Lastly, Bercea et al [6] provided a 3−approximation algorithm when F = C and 5−approximation when F = C for Problem 2. However, the LP uses Θ(N 2 ) LP variables and constraints, and does not allow a point to belong to multiple groups, i.e ∆ = 1.…”

Section: Preliminariesmentioning

confidence: 99%

“…This ensures that no group "dominates" any cluster. However, we note that [2] and [6] did not allow overlapping between groups, e.g. an input point cannot belong to both "male" and "Asian" groups.…”

Section: Introduction and Prior Workmentioning

confidence: 99%

“…However, their definition of fairness is quite restrictive, as sometimes there are some input instances where such a fair cluster by that definition does not exist. [6]. Their idea of fairness stems from the classical k−center problem.…”

Section: Introduction and Prior Workmentioning

confidence: 99%

“…Note that [2] formulation is a special case of [5] formulation. [6] Also gave an elegant, yet computationally expensive, approximation framework for essentially fair clustering for various objectives including k−center.…”

Section: Introduction and Prior Workmentioning

confidence: 99%

See 4 more Smart Citations

KFC: A Scalable Approximation Algorithm for $k$-center Fair Clustering

Harb¹,

Lam²

2020

Preprint

View full text Add to dashboard Cite

In this paper, we study the problem of fair clustering on the k−center objective. In fair clustering, the input is N points, each belonging to at least one of l protected groups, e.g. male, female, Asian, Hispanic. The objective is to cluster the N points into k clusters to minimize a classical clustering objective function. However, there is an additional constraint that each cluster needs to be fair, under some notion of fairness. This ensures that no group is either "over-represented" or "under-represented" in any cluster. Our work builds on the work of Chierichetti et al. (NIPS 2017), Bera et al. (NeurIPS 2019), Ahmadian et al. (KDD 2019), and Bercea et al. (APPROX 2019). We obtain a randomized 3−approximation algorithm for the k−center objective function, beating the previous state of the art (4−approximation). We test our algorithm on real datasets, and show that our algorithm is effective in finding good clusters without over-representation or underrepresentation, surpassing the current state of the art in runtime speed, clustering cost, while achieving similar fairness violations. * Both authors contributed equally to the paper. Author order is in alphabetical order. We thank Ho Chung Leon Law from University of Oxford for recommending the algorithm name. We also thank all the reviewers for their incisive comments and helping us improve the paper.34th Conference on Neural Information Processing Systems (NeurIPS 2020),

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Preliminariesmentioning

confidence: 99%