Interpretability and causality have been acknowledged as key ingredients to the success and evolution of modern machine learning systems. Graphical models, and more specifically directed acyclic graphs (DAGs, also known as Bayesian networks), are an established tool for learning and representing interpretable causal models. Unfortunately, estimating the structure of DAGs from data is a notoriously difficult problem, and as a result existing approaches typically come with strong assumptions such as linearity or faithfulness. We will discuss our recent work towards overcoming these barriers by re-formulating the problem in more “user-friendly” ways based on continuous optimization and standard nonparametric regression techniques. A nice byproduct of this approach is a provably polynomial-time algorithm with finite-sample guarantees. The result is a general framework for learning parametric and nonparametric DAG models with rigourous theoretical guarantees.
Bio: Bryon Aragam is an Assistant Professor of Econometrics and Statistics in the Booth School of Business at the University of Chicago. His research interests include statistical machine learning, unsupervised learning (graphical models, representation learning, latent variable models, etc.), nonparametric statistics, and causal inference. He is also involved with developing open-source software and solving problems in interpretability, ethics, and fairness in artificial intelligence. Prior to joining the University of Chicago, he was a project scientist and postdoctoral researcher in the Machine Learning Department at Carnegie Mellon University. He completed his PhD in Statistics and a Masters in Applied Mathematics at UCLA, where he was an NSF graduate research fellow.