• Corpus ID: 232105026

Approximation Algorithms for Socially Fair Clustering

@inproceedings{Makarychev2021ApproximationAF,
  title={Approximation Algorithms for Socially Fair Clustering},
  author={Yury Makarychev and A. Vakilian},
  booktitle={COLT},
  year={2021}
}
We present an (e log l log log l )-approximation algorithm for socially fair clustering with the lpobjective. In this problem, we are given a set of points in a metric space. Each point belongs to one (or several) of l groups. The goal is to find a k-medians, k-means, or, more generally, lpclustering that is simultaneously good for all of the groups. More precisely, we need to find a set of k centers C so as to minimize the maximum over all groups j of ∑ u in group j d(u,C) . The socially fair… 
Improved Approximation Algorithms for Individually Fair Clustering
TLDR
The approach suggests a reduction from an individually fair clustering to a clustering with a group fairness re-quirement proposed by Kleindessner et al.
Tight FPT Approximation for Socially Fair Clustering
TLDR
The goal in the socially fair k-median problem is to find a set C ⊆ F of k centers that minimizes the maximum average cost over all the groups.
Approximating Fair Clustering with Cascaded Norm Objectives
TLDR
This work utilizes convex programming techniques to approximate the (p, q)-Fair Clustering problem for different values of p and q and gets an O(k), which nearly matches a k lower bound based on conjectured hardness of Min k-Union and other problems.
Constant-Factor Approximation Algorithms for Socially Fair k-Clustering
TLDR
The performance of these algorithms are compared with existing bicriteria algorithms as well as exactly k center approximation algorithms on benchmark datasets, and it is found that these algorithms also outperform existing methods in practice.
Fair Representation Clustering with Several Protected Classes
TLDR
An O(log k)-approximation algorithm that runs in time nO(ℓ) that fairly represents all groups and minimizes the ℓ1-objective ∑x ∈ Xd(x, ϕ(x)).
FPT Approximation for Fair Minimum-Load Clustering
TLDR
This work makes substantial progress in understanding the structure of the minimum-load objective, advancing the frontiers of knowledge about the MLkC problem and achieving the first constant approximations for this problem in general and Euclidean metric spaces.
Approximate the individually fair k-center with outliers
TLDR
A 4-approximation algorithm is presented, based on which an improved algorithm is developed from a practical perspective, so as to minimize the maximum fairness ratio over all served vertices.
Efficient Algorithms For Fair Clustering with a New Fairness Notion
TLDR
A new notion of fairness is proposed, which is called τ -ratio fairness, that strictly generalizes the Balance property and enables a finegrained efficiency vs. fairness trade-off.
On Coresets for Fair Regression and Individually Fair Clustering
TLDR
This paper defines coresets for Fair Regression with Statistical Parity (SP) constraints and for Individually Fair Clustering and shows that to obtain such coresets, it is sufficient to sample points based on the probabilities dependent on combination of sensitivity score and a carefully chosen term according to the fairness constraints.
Towards Fair Unsupervised Learning
TLDR
The concept of fairness functional is introduced that encompasses both traditional unsupervised learning techniques and min-max algorithms (whereby one minimises the maximum group loss) and is designed to design straightforward alternate convex search or biconvex gradient descent algorithms that also provide partial debiasing techniques.
...
...

References

SHOWING 1-10 OF 49 REFERENCES
A Constant-Factor Approximation Algorithm for the k-Median Problem
TLDR
This work presents the first constant-factor approximation algorithm for the metric k-median problem, and improves upon the best previously known result of O(log k log log log k), which was obtained by refining and derandomizing a randomized O( log n log log n)-approximation algorithm of Bartal.
Proportionally Fair Clustering Revisited
TLDR
This work derives generalization bounds which show that an approximately proportionally fair clustering for a large number of points can be learned from a small number of samples, and shows that for the L2 distance, checking whether a proportionallyFair clustering exists and implementing greedy capture over an infinite metric space are NP-hard problems, but (approximately) solvable in special cases.
The matroid median problem
TLDR
A constant factor approximation algorithm is given for the matroid median problem and it is shown that this second phase LP is in fact integral; the integrality proof is based on a connection to matroid intersection.
Fair Algorithms for Clustering
TLDR
This work significantly generalizes the seminal work of Chierichetti this http URL and transforms any vanilla clustering solution into a fair one incurring only a slight loss in quality.
Socially Fair k-Means Clustering
TLDR
It is found that on benchmark datasets, Fair-Lloyd exhibits unbiased performance by ensuring that all groups have equal costs in the output k-clustering, while incurring a negligible increase in running time, thus making it a viable fair option wherever k-means is currently used.
Constant approximation for k-median and k-means with outliers via iterative rounding
TLDR
A new iterative rounding framework for many clustering problems is presented, and an α1- and (α1 + є)-approximation algorithms for matroid and knapsack median problems respectively are given, improving upon the previous best approximations ratios.
Fair k-Centers via Maximum Matching
TLDR
This paper combines the best of each algorithm by presenting a linear-time algorithm with a guaranteed 3-approximation factor and provides empirical evidence of both the algorithm’s runtime and effectiveness.
On the cost of essentially fair clusterings
TLDR
A relaxed fairness notion under which bicriteria constant-factor approximations for all of the classical clustering objectives are given, which can be established belatedly, in a situation where the centers are already fixed.
Clustering without Over-Representation
TLDR
This paper obtains an algorithm that has provable guarantees of performance and a simpler combinatorial algorithm for the special case of the problem where no color has an absolute majority in any cluster.
Sampling a Near Neighbor in High Dimensions — Who is the Fairest of Them All?
TLDR
This work shows that LSH based algorithms can be made fair, without a significant loss in efficiency, and develops a data structure for fair similarity search under inner product that requires nearly-linear space and exploits locality sensitive filters.
...
...