New Seminar on Causal Discovery to be offered by CMU in Fall, 2015

A new seminar on causal discovery is to be offered by Carnegie Mellon University in Fall, 2015 (Course #80-516 for undergraduate level;  Course #80-816 for graduate level;  Section A for both levels).  The instructor will be Kun Zhang, a new faculty member in CMU’s Philosophy Department.  Pitt students can register now through the Pitt cross-registration process.

Click here for the announcement and syllabus

Description

Causal connections are usually more interesting or helpful than purely associational information. This course is mainly concerned with systematic approaches to discovering causal connections from data in various scenarios and the question why causation plays an important role in science, i.e., how it is helpful in understanding,  decision making, and prediction in complex environments.

We will study the difference between causal and non-causal systems and make an attempt to characterize a causal system. Apart from identification of causal effects, we will explore two causality-related areas. One is causal discovery, i.e., going beyond the observational data to the underlying causal information. It is well known that “correlation does not imply causality,” but we will make this statement more precise by asking what information in the data and what assumptions enable us to discover causal information from purely observed data. This will cover constraint-based causal discovery, causal discovery based on structural equation models, causal discovery from time series, difficulties in practical causal discovery, causality in neuroscience, causality in biology, and causality in economics and finance. More importantly, we will have the opportunity to solve problems in various fields from a causal perspective: participants may bring any causal problems they are interested in, and we will work together to find potential solutions. The other is how to properly make use of causal information. This includes counterfactual reasoning, improving machine learning in light of causal knowledge, and forecasting in nonstationary unseen environments.

Overall, this course aims to provide fundamentals of causal discovery and inference, review emerging methods for causal discovery, report their applications, find practical causal problems in various fields, and work out potential solutions.