Abstract

We show that complex-valued neural networks with the modReLU activationfunction $\sigma(z) = \mathrm{ReLU}(|z| - 1) \cdot z / |z|$ can uniformlyapproximate complex-valued functions of regularity $C^n$ on compact subsets of$\mathbb{C}^d$, giving explicit bounds on the approximation rate.

 

Quick Read (beta)