This course combines lectures and seminar-style discussions to understand the fast-moving intersection of causality and machine learning (ML). The study of causality has long propelled fields like epidemiology, medicine, and economics. More recently, ML and causality are increasingly connected. On one hand, Parameterizing causal models with neural networks offers attractive out-of-distribution (OOD) guarantees as well as new ways of causally analyzing complex data. On the other hand, causal notions of intervening and inferring effects offer ways of analyzing mechanisms in large models. This course covers background on causal models, and recent results on core topics like causal representation learning, causal structure discovery, causal abstraction and uses of causality with large pre-trained models. The course will consist of readings (with associated write-ups) and a project.

I will assume programming experience and familiarity with topics taught in Fundamentals of Machine learning (or equivalent).  Background on deep learning, unsupervised machine learning, and probabilistic graphical models will be very useful.