The official syllabus is at https://github.com/csquires/6.S091-causality/blob/main/syllabus.pdf.
Lecture notes and links to recordings will be posted on this page.
Instructor: Chandler Squires
TA: Katie Matton
Time: Tuesday and Thursday, 1-3pm
Dates: 01/10/23 - 02/02/23
Location: 4-231
Credit: 6 units
In this course, we will cover introductory material from three active research areas related to causality and machine learning. In the first third of the course, we will discuss the fundamentals of policy evaluation, where a known causal structure is used to estimate causal quantities such as (conditional) average treatment effects. In this section, we will cover algorithms for identification of causal estimands, as well the principles behind state-of-the-art estimation methods based on double/de-biased machine learning. In the second third of the course, we will consider causal structure learning, i.e., the estimation of an unknown causal structure from data. We will cover classical algorithms such as the PC algorithm, as well as newer methods which incorporate interventional data and allow for unobserved confounding. We will also cover experimental design techniques for causal structure learning. In the final third of the course, we will discuss the emerging field of causal representation learning, highlighting recent papers which connect machine learning with more traditional causal principles.
Tuesday, Jan 10: Introduction to Structural Causal Models (lecture notes, recording)
Thursday, Jan 12: Policy Evaluation I: Identification (lecture notes, recording)
Tuesday, Jan 17: Policy Evaluation II: Estimation
Thursday, Jan 19: Causal Structure Learning I: Identifiability and Classical Approaches
Tuesday, Jan 24: Causal Structure Learning II: Gradient-based Approaches
Thursday, Jan 26: Causal Structure Learning III: Experimental Design
Tuesday, Jan 31: Causal Representation Learning I: Learning Latent Causal Structure
Thursday, Feb 2: Causal Representation Learning II: Domain Generalization
- Problem sets must be done in LaTeX
- Printed problem sets must be turned in at the beginning of lecture.
- Due dates:
Thursday, Jan 19: PSet 1 due at 1pm EST
Thursday, Jan 26: PSet 2 due at 1pm EST
Thursday, Feb 2: PSet 3 due at 1pm EST