Department of Computer Science & Engineering

Description: Current successes in machine learning build on correlation and not causation. There is a well-developed theory of causality and causal reasoning which is recently getting renewed attention. This course introduces the concepts and algorithms of causal inference enabling students to answer standard types of queries, including associational, causal, and counterfactual. Background in introductory probability and statistics is essential.

Course Content : Introduction: concepts in probability; graphs; Bayesian networks; d-separation; causal Bayesian net-works; functional causal models; Causal Diagrams and Identification: interventions; confounding bias; back-door criteria; front door criteria; do-calculus; Actions and Sequential Plans: conditional actions and stochastic policies; Galles and Pearl condition for identification; closed form expression for identification; dynamic plans; sequential back door criteria; g-identifiability; direct vs. total effects; Simpson Paradox: problem definition, sure-thing principal; causal and associational definitions for no-confounding; failure of associational criteria; stable vs incidental unbiasedness; collapsibility, ex-changeability and confounding. Causality in Social Sciences: structural equation models(SEM); testable implications of structural models; model equivalence; graphs and identifiability; interventional interpretation of SEM; Counterfactuals: structural model semantics; definitions; deterministic analysis; probabilistic analysis; twin network method; applications; axiomatic characterization;

TextBooks : 1. Judea Pearl, Causality: Models, Reasoning and Inference, Cambridge University Press, 2nd edition, 2009. 2. Hernan MA, Robins JM, Causal Inference, Boca Raton: Chapman and Hall, CRC, 2019

Reference Books : 1. Tian J, Studies in Causal Reasoning and Learning, PhD Thesis, UCLA, 2002 2. Chen. B. and Pearl. J, Graphical tools for linear structural equation modeling. Technical Re-port-432, UCLA, 2015. 3. Morgan, S., & Winship, C., Counterfactuals and Causal Inference: Methods and Principles for Social Research (Analytical Methods for Social Research). Cambridge: Cambridge University Press, 2014. 4. Research Papers.

Prerequisite: MA 2040 OR CS6015 AND CS6380 OR CS5691