Embarking on Transparent and Fair Journeys in Data Science by Lara Kassab, UCLA

Math Colloquium by Lara Kassab, UCLA, Monday January 22, 1:00 – 2:00pm, North Science Building 015, Wachenheim

Abstract: Machine learning, particularly automated decision-making, plays a significant role in numerous sensitive aspects of our society. In this talk, we touch on the challenges of transparency and fairness in machine learning. We introduce interpretable linear models for learning tasks based on matrix and tensor factorizations. These models have versatile applications and leverage inherent structures in the data, such as nonnegativity and multiway structures. Last, we present ongoing work on fairness-aware dimension reduction techniques, particularly using nonnegative matrix factorizations.