Tianxi Cai

ATLAS achieves near-oracle performance in latent factor regression by effectively disentangling invariant and heterogeneous factors, even in the presence of diverse environments.

Reduce testing costs without compromising predictive accuracy by learning cost-optimal sequential decision policies from retrospective data, even with informative missingness.

By injecting semantic knowledge into latent space models, this work shows how to learn more robust representations from imbalanced data, a pervasive challenge in fields like healthcare.