Causality-inspired Distributional Robustness
Robust, reliable and interpretable machine learning is a big emerging theme in data science. New mathematical connections between distributional robustness and causality provide methodological paths for improving the reliability and understanding of statistical machine learning algorithms, with wide-ranging prospects for various applications. After introducing the fundamental concepts, we discuss recent progress on Distributional Robustness via Invariant Gradients (DRIG), Invariant Probabilistic Prediction (IPP), and practical domain adaptation for predicting patient status in new ("out-of-distribution") intensive care units.