Title: Causal Discovery from Observational Data in the Presence of Latent Confounders and Complex Data

Date: Monday, July 22

Time: 11:00 AM ET

Location: Coda C1115 Druid Hills & Zoom

Yuqin Yang

Machine Learning Ph.D. Student

School of Mathematics
Georgia Institute of Technology

Committee

1. Prof. Mark Davenport (Advisor)

2. Prof. Negar Kiyavash (Co-advisor)

3. Prof. Yao Xie

4. Prof. Ashwin Pananjady

5. Prof. Kun Zhang

Abstract

Causal discovery aims to recover causal relationships among variables of interest from only observational data. However, there are complexities in real-life data that make causal discovery even more challenging. Some of the main sources of data complexity include latent confounding, deterministic relations, measurement error, and data heterogeneity. The majority of causal discovery methods assume that these complexities are absent in the system. Naturally, naive applications of these approaches to settings that indeed are subject to data complexity issues lead to detecting spurious or erroneous causal links among variables of interest.

The focus of the dissertation is on developing causal discovery methods that are capable of handling these data complexities. Specifically: (1) We study the problem of causal discovery in linear causal models with deterministic relations and latent confounding. (2) We study the problem of causal discovery in linear causal models in the presence of latent confounding and/or measurement error. (3) We study the problem of learning the unknown intervention targets in linear or nonlinear causal models from a collection of interventional data obtained from multiple environments, both under causal sufficiency assumption and in the presence of latent confounding.