80-314 Causal Discovery, Statistics, and Machine Learning
9 units, Carnegie Mellon University
Statistics and Machine Learning have made tremendous strides in recent years in solving a wide variety of regression and classifiction problems. However, causal discovery problems i.e. discovery of which variables are affected by a given variable that undergoes change due to an external intervention, either manmade or natural, and to what extent other variables are affected by such a change are distinct and more difficult problems. Causal discovery problems arise not only in scientific contexts e.g. discovering which genes regulate which other genes but also in some machine learning contexts e.g. transfer learning problems. This course will i describe how causal discovery problems differs from regression and classification problems in goals, methods, and fundamental assumptions, ii descibe recent advances in modifying machine learning and statistical algorithms to deal with causal discovery problems involving such difficulties as latent confounders, measurement error, selection bias, etc., and iii what the outstanding problems in causal discovery are and future directions the field might take. Students should have taken at least one course in statistics or machine learning, or obtain the permission of the instructor.
Prerequisites: none listed
Courseweave needs JavaScript to show sections, workloads and planning tools.