AI Institute for Societal Decision Making Seminar
— 4:00pm
Location:
In Person and Virtual - ET
-
Newell-Simon 4305 and Zoom
Speaker:
KUN ZHANG
,
Associate Professor , Department of Philosophy, and, Affiliate Faculty Member, Machine Learning Department, Carnegie Mellon University
https://www.andrew.cmu.edu/user/kunz1/
Causal Representation Learning and its Connection with Decision Making
Causal representation learning aims to reveal the underlying high-level hidden causal variables and causal influences. It can be seen as a special case of causal discovery, whose goal is to recover the underlying causal structure or causal model from observational data. The modularity property of a causal system implies properties of minimal changes and independent changes of causal representations; causal models naturally support decision making in three ways. First, they tell us which variables are true causal variables that can help make a difference. Second, they allow us to foresee the consequences of different actions and figure out which interventions are effective. Third, they help us quickly adapt to new environments in the presence of nonstationarity, by identifying minimal changes across environments.
In this talk, we first show how to recover the underlying causal representations from observational data with identifiability guarantees: under appropriate assumptions, the learned representations are consistent with the underlying causal process. Various problem settings are considered, involving independent and identically distributed (i.i.d.) data, temporal data, or data with distribution shift as input. We then demonstrate when identifiable causal representation learning can benefit from flexible deep learning and when suitable parametric assumptions have to be imposed on the causal process. Finally, we show how to learn compact, interpretable state representations, together with the underlying causal model, in deep reinforcement learning and how to make quick adaptations in transfer reinforcement learning.
—
Kun Zhang is an associate professor in the CMU philosophy department and an affiliate faculty member in the machine learning department. His research interests lie in machine learning and artificial intelligence, especially in causal discovery and causality-based learning. He develops methods for automated causal discovery from various kinds of data, investigates learning problems including transfer learning and deep learning from a causal view, and studies philosophical foundations of causation and machine learning. On the application side, he is interested in neuroscience, computational finance, and climate analysis.
Refreshments will be available In Person and Zoom Participation. See announcement.