Group Fair Clustering Revisited -- Notions and Efficient Algorithm

Shivam Gupta (Indian Institute of Technology Ropar), Ganesh Ghalme (Indian Institute of Technology Hyderabad), Narayanan C. Krishnan (Indian Institute of Technology Palakkad), Shweta Jain (Indian Institute of Technology Ropar)

Abstract

This paper considers the problem of group fairness in clustering. We propose a new fairness notion which strictly generalizes existing notions, and we theoretically analyze the relationships between several existing notions. Finally, we propose a simple and efficient greedy round-robin-based algorithm (FRAC 𝑂𝐸) and extensive experiments to validate its efficacy across multiple datasets.