Thursday, October 12th at UIC
About: Bruno Ribeiro is an Associate Professor in the Department of Computer Science at Purdue University and currently a Visiting Associate Professor at Stanford University. Before joining Purdue, he earned his Ph.D. from the University of Massachusetts Amherst and was a postdoctoral fellow at Carnegie Mellon University. Ribeiro has made significant contributions in the intersection between invariant theory, graphs and machine learning. Ribeiro received an NSF CAREER award in 2020, an Amazon Research Award in 2022, and multiple best paper awards.
Title and Abstract
Title: Insights into Causal Link Prediction through Causal Lifting
Abstract: In this talk we explore the role of symmetries in predicting the evolution of links on graphs both with and without interventions. Specifically, we will see that standard (associational) temporal link prediction tasks can always be solved by static graph machine learning methods. Further, using invariant theory, we will highlight the limitations of matrix factorizations as graph embeddings when predicting intervention outcomes on links within path-dependent graphs (e.g., friendship recommendations in a social network and product recommendations in e-commerce). Finally, I will introduce the symmetry-based concept of “Causal Lifting” for predicting the effect of link interventions on path-dependent graphs and discuss some applications.
Logistics
- Date: Thursday, October 12th, 11:00AM
- In-person Location: SEO 1000 at UIC