Abstract:This is essentially the transcript of the Peter Henrici Prize Lecture, given on July 15 at ICIAM 2019 in Valencia. The talk was designed to give a broad overview of • some of the current work on integrating machine learning with scientific modeling (e.g. physicsbased modeling) to address some of the most challenging problems in a variety of disciplines, and • some of the current work on building amathematical theory of machine learning. There were two basic messages that I wanted to convey. First, in theoretical and computational science and engineering, a fundamental obstacle that we have encountered is our limited ability to deal with problems in high dimension. Machine learning has now provided new tools for

Machine Learning: Mathematical Theory and Scientific Applications