Bias Mitigation in the Wild: Challenges and Opportunities

Decanato - Facoltà di scienze informatiche

Data: 2 Ottobre 2024 / 14:30 - 15:30

USI East Campus, Room D1.02

The seminar is co-organised by IDSIA USI-SUPSI and the Computer Systems Institute at USI Faculty of Informatics. This is the first initiative promoted by the Swiss-XAI-Cluster, a recently-born and expanding network of junior researchers with interest in XAI and surrounding areas located in Swiss universities (www.swissxai.ch)

Speaker: Mateo Espinoza Zarlenga, University of Cambridge

Abstract: Deep neural networks trained via empirical risk minimisation often exhibit significant performance disparities across groups, particularly when group and task labels are spuriously correlated (e.g., "grassy background" and "cows"). In this talk, I will first argue that previously proposed bias mitigation methods that aim to address this issue often either rely on group labels for training or validation, or require an extensive hyperparameter search, even when this assumption is not explicitly made. Such data and computational requirements hinder the practical deployment of these methods, especially when datasets are too large to be group-annotated, computational resources are limited, and models are trained through already complex pipelines. With this in mind, I will outline some of the challenges that may need to be addressed to design practical bias mitigation methods. Then, I will describe Targeted Augmentations for Bias mitigation (TAB), a new approach that takes these design principles into consideration. I will conclude by showing how TAB, a simple hyperparameter-free framework that leverages the entire training history of a helper model to identify spurious samples, improves worst-group performance without any group information or model selection.

Biography: Mateo Espinoza Zarlenga is a final-year PhD student and Gates scholar at the University of Cambridge. His research is focused on developing interpretable deep learning methods that explain their predictions using human-understandable concept representations. Mateo has published papers at NeurIPS, ICML, AAAI, AIES, TMLR, ECCV, and ASPLOS. He holds an MPhil from the University of Cambridge and an MEng and BA from Cornell University.

Facoltà