Webinar video https://www.youtube.com/watch?v=x9LOH06YBFA
Presentation slides
Abstract
As the adoption of machine learning models and automated decision-making systems grows, understanding their behavior is becoming more and more critical to ensure their reliability and fairness. Current approaches predominantly focus on characterizing classification performance either considering a single instance, or the entire model. However, these approaches provide no indication if differences in the model behavior exist across subsets of data. For example, speech models may struggle with performance inconsistencies across different population subgroups, leading to degraded accuracy for certain speaker demographics, accents, or recording conditions. These discrepancies may originate from multiple reasons, such as imbalanced training data, suboptimal representation learning, and limitations in model generalization.
In this seminar, we will introduce the notion of subgroup and explore a variety of domains, such as model bias mitigation and model drift detection, in which subgroup discovery may improve model robustness and reliability in real-world applications.
Short Bio
Elena Baralis is full professor at the Department of Control and Computer Engineering of the Politecnico di Torino, Italy since January 2005. She is currently the Deputy Rector of the Politecnico di Torino. Her current research interests are in the field of machine learning and data science, more specifically on explainable AI, bias detection in data analytics, and machine learning algorithms for big data. She has published over 200 papers in international journals and conference proceedings. She was PI for several European and national research projects on data science and machine learning. She was general co-chair of the 30th ECML/PKDD in 2023 and program co-chair of IEEE ICDM 2024. She has served as area chair and in the program committees of international conferences and workshops, among which IEEE ICDM, ACM SIGMOD, VLDB, ACM SAC, ACM CIKM, ECML/PKDD.
Information at FEP News
More information at News U. Porto