Lectures on Causality by Karthikeyan Shanmugam

A lecture series on causal discovery and inference at the CSA Department, IISc Bangalore.

Location: CSA Seminar Hall (Room 254, CSA Building, IISc)

Dates: Mondays and Thursdays 21st September to 9th October

Timing: 4.00pm-05.30pm

Speaker Website: https://sites.google.com/corp/view/karthikeyan-shanmugam

Speaker Affiliation: Google Research, Bangalore

Schedule for Talks:

September 21 (Thursday).

First Lecture – Sure thing principle and its causal variant. Simpson’s paradox – main motivating central issue in causal inference. Formal Setting of potential outcomes model. Confounder balancing under ignorability, Rubin Rosenbaum results on importance weighing and stratification.

September 25 (Monday).

Second Lecture – Various estimators for ATE (Average Treatment Effect) under ignorability continued. CATE estimation and estimators for it. Individual treatment effect estimators – X learner and T learner. Representation learning for ITE – CFR Net, TAR net

October 5 (Thursday).

Third Lecture – Continue representation learning for treatment effect (state some open problems). Introduction to Sensitivity Analysis. Instrumental Variable approach for linear models (if time permits).

October 9 (Monday).

Fourth Lecture – Looking back at Simpson’s paradox – what is ignorability does not hold. Intro to Pearlian Approach to Causality.

Definition of Structural Equation Models, Causal Bayesian Networks. Do calculus definitions. Equivalence to the modularity or interventional invariance condition.

References:

(i) Hernán MA, Robins JM. Causal Inference: What If.

(ii) Pearl. Causality.

Speaker Bio:

Karthikeyan Shanmugam is a Research Scientist at Google Research India (Bengaluru) since April 2022. He is a part of the Machine Learning Foundations and Optimization Team. Previously, he was a Member of Research Staff with IBM Research AI, NY and a Herman Goldstine Postdoctoral Fellow at IBM Research, NY. He obtained his Ph.D. from UT Austin where his advisor was Alex Dimakis. He obtained his master’s degree from the University of Southern California, and prior to that his B.Tech (Dual Degree) from IIT Madras.

His research interests broadly lie in Graph algorithms, Machine learning, Optimization, Coding Theory, and Information Theory. Specifically in machine learning, my recent focus is on Causal Inference, Bandits/RL and Explainable AI.

All are invited. Snacks and Tea will be served post the talks.