
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.


coremsg

learning-efficient-approximations-of-high-dimensional-smooth-functions-from-limited-data


Learning Efficient Approximations of High-Dimensional Smooth Functions from Limited Data


title_rewrite


The authors propose two algorithms that utilize the ANOVA decomposition to learn low-order functions with few variable interactions, enabling reliable identification of important input variables and their interactions. This approach improves the interpretability of existing random Fourier feature models.


anova-boosting-for-random-fourier-features-an-interpretable-approach-to-high-dimensional-function-approximation


ANOVA-Boosting for Random Fourier Features: An Interpretable Approach to High-Dimensional Function Approximation
