The discovery of causal relationships is of utmost importance in science and engineering, and causal inference, particularly using observational data, has recently become a popular topic in machine learning. While there is no widely agreed-upon definition for causal strength, it’s intuitive that the strength of a causal relationship is important for decision-making, and may contain important information for learning algorithms. In this talk, we’ll discuss one way to quantify causal influences called the differential causal effect (DCE), and explore two recent approaches to causal inference using the DCE. The first is an online method for detecting hidden confounding variables in time series, the presence of which complicates causal inference significantly. The second is an interpretable continuous optimization method for the discovery of causal relationships, where causal strength is used and regularized in training.
Daniel Waxman received the B.S. degree in mathematics and applied mathematics & statistics from Stony Brook University, Stony Brook, NY, USA in 2021. He is currently working towards the Ph.D. degree in electrical engineering at Stony Brook University with Petar Djuric. His research interests include Bayesian machine learning, statistics, and causal structure learning.