Physics Colloquium: Trustworthy Machine Learning for Experimental Design

Prof. Angie Liu, Johns Hopkins University

Abstract: The unprecedented prediction accuracy of modern machine learning beckons for its application in a wide range of real-world applications, including autonomous robots, healthcare, scientific experimental design, and many others. A key challenge in such real-world applications is that the test cases are not well represented by the pre-collected training data. To properly leverage learning in such domains, we must go beyond the conventional learning paradigm of maximizing average prediction accuracy with generalization guarantees that rely on strong distributional relationships between training and test examples.

In this talk, I will describe machine learning and deep learning frameworks that offer rigorous guarantees under data distribution shift and introduce how to integrate them into real-world systems to guide decision-making. I will showcase the practicality of them in applications on exploration-based data collection and experimental design. I will also introduce a survey of other real-world applications, including applications in material sciences, that would benefit from this framework for future work.