Learning Efficient Approximations of High-Dimensional Smooth Functions from Limited Data
Functions that are holomorphic in high-dimensional parameter spaces can be efficiently approximated using sparse polynomial expansions or deep neural networks, even when only limited training data is available.