Quaternion neural networks for multidimensional applications. An Overview

Although the concept of quaternion neural networks (QNNs) is not entirely new, interest in this type of network has increased recently. The possibilities they offer for effectively processing datasets characterised by inherent multidimensionality make them an interesting object of study and investigation. This contribution aims at highlighting the relevant key principles underlying these adaptive architectures and at providing an essential overview of the different classes and application areas available, serving as a tutorial introduction to the world of hypercomplex neural networks.