CS Seminar: Murat Kocaoglu

Description

Murat Kocaoglu, an assistant professor in the School of Electrical and Computer Engineering at Purdue University, will give a talk titled "Causal Machine Learning: Fundamentals and Applications" for the Department of Computer Science.

Refreshments are available starting at 10:30 a.m. The seminar will begin at 10:45 a.m.

Abstract:

Causal knowledge is central to solving complex decision-making problems across engineering, medicine, and cyber-physical systems. Causal inference has been identified as a key capability to improve machine learning systems' explainability, trustworthiness, and generalization. After a brief introduction to causal modeling, this talk explores two key problems in causal ML. In the first part of the talk, we will focus on the problem of root-cause analysis (RCA), which aims to identify the source of failure in large, modern computer systems. We will show that by leveraging ideas from causal discovery, it is possible to automate and efficiently solve the RCA problem by systematically using invariance tests on normal and anomalous data. In the second part of the talk, we consider causal inference problems in the presence of high dimensional variables, e.g., image data. We show how deep generative models, such as generative adversarial networks and diffusion models, can be used to obtain a representation of the causal system and help solve complex, high-dimensional causal inference problems. This approach enables both causal invariant prediction and evaluation of black box conditional generative models.