Publication Date

2025

Document Type

Dissertation/Thesis

First Advisor

Krishtal, Ilya

Degree Name

M.S. (Master of Science)

Legacy Department

Department of Mathematical Sciences

Abstract

Neural networks have become central to modern natural science. One of the key components of these networks is the choice of a suitable activation function. In this manuscript, we explore the selection of an optimal activation function in the context of complex-valued neural networks. We aim to characterize the activation functions $\sigma\colon \mathbb{C} \rightarrow \mathbb{C}$, where each neuron performs the operation $\mathbb{C}^N \rightarrow \mathbb{C}$, defined as: $z \mapsto \sigma(b + w^Tz)$. First, we briefly discuss the real-valued counterpart, including the importance of function approximation and the renowned universal approximation theorem. Subsequently, we examine various criteria for complex activation functions that provide a universal network class for shallow and deep networks.

Extent

48 pages

Language

en

Publisher

Northern Illinois University

Rights Statement

In Copyright

Rights Statement 2

NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.

Media Type

Text

Share

COinS