Analytic function approximation by path norm regularized deep networks
Aleksandr Beknazaryan · arXiv (Cornell University) · 2021
We show that neural networks with absolute value activation function and with the path norm, the depth, the width and the network weights having logarithmic dependence on $1/\varepsilon$ can $\varepsilon$-approximate functions that are analytic on certain regions of $\mathbb{C}^d$.