Clustering is a classical unsupervised machine learning technique. It has various applications in criminal justice, automated resume processing, bank loan approvals, recommender systems, and many more. Despite being so popular, traditional clustering algorithms may result in discriminatory behavior towards a group of people (or individuals) and have societal impacts. It has led to the study of fair clustering algorithms that aim to minimize the clustering cost while ensuring fairness criteria. In the talk, we will primarily focus on group fair clustering which primarily requires that data points from each protected group have approximately equal representation in every cluster. This talk will start by discussing the relationships between existing notions in group fair clustering and algorithms in this literature. I will also discuss the advantages and disadvantages of existing algorithms in terms of theoretical guarantees, time complexity, and reproducibility. Finally, the talk will conclude by providing new directions and open problems in fair clustering.
Shweta Jain, IIT Ropar
Shweta Jain is an assistant professor in the computer science and engineering department at the Indian Institute of Technology, Ropar. Her research interests are machine learning, game theory, and mechanism design. She obtained her Ph.D. in Computer Science and Engineering from IISc, Bangalore, and worked as an assistant professor at IIT Bhubaneshwar before joining IIT Ropar. She is also heading the Indo-Taiwan Joint Research Center at IIT Ropar.