3.0

Average Course Rating

Data often provides a projection of the inner workings of real-world systems. Many problems require a deeper understanding of the cause-and-effect relations that underlie the data-generating process. Causal discovery refers to the process of learning a graphical representation of these causal relations, called a causal graph. Such representations can be used for causal effect estimation, as well as other applications, such as system monitoring and diagnosis (e.g., root-cause analysis). In this course, we will learn the foundational principles behind causal discovery. We will explore the fundamental limits of causal discovery under well-defined assumptions and discuss algorithmic approaches for learning causal graphs from data. Required Course Background: a graduate level course in machine learning or basics of data structures, linear algebra and probability. Students may receive credit for at most one of 601.478/678.