Deep neural network approximation of composite functions without the curse of dimensionality

Adrian Riekert

In this article we identify a general class of high-dimensional continuous functions that can be approximated by deep neural networks (DNNs) with the rectified linear unit (ReLU) a…