CS seminar: The utility-fairness trade-offs in learning fair representation
This event is in the past.
11:30 a.m. to 12:20 p.m.
Speaker
Dr. Vishnu Boddeti, Associate Professor, Michigan State University
Abstract
Machine learning systems are increasingly being employed to render or support automated decisions in many contexts, including search engines, law enforcement, healthcare, and many more. There is growing evidence that such systems propagate and amplify any biases present in the data, leading to discriminatory outcomes for subgroups in a population. Learning representations of data that maintain their utility while suppressing information that could lead to biased predictions by downstream tasks is the dominant approach to mitigate the unfairness of machine learning systems. Solutions to fair representation learning problems may lead to trade-offs between utility and fairness. In this talk, I will address the following questions about such trade-offs.
Q1. What are the different types of trade-offs between utility and fairness?
Q2. Under what circumstances is there a trade-off between utility and fairness?
Q4. How can we characterize the trade-offs?
Q3. How far are the existing fair representation learning algorithms in achieving the optimal trade-offs?
Biography
Vishnu Naresh Boddeti is an Associate Professor in the Department of Computer Science and Engineering at Michigan State University. He received a B.Tech degree in Electrical Engineering at the Indian Institute of Technology, Madras and a Ph.D. degree in the Electrical and Computer Engineering program at Carnegie Mellon University in 2013. His research interests are Computer Vision, Pattern Recognition, and Machine Learning. His current focus is on building efficient and trustworthy machine learning systems, including algorithmic fairness, security, and privacy. Papers co-authored by him have received the Best Paper Awards at BTAS 2013 and GECCO 2019 and the Best Student Paper Awards at ACCV 2018, SMAIS 2022, IJCB 2022, and TBIOM 2023.