Hypercomplex-valued Neural Networks for Signal Processings

MLSP 2023 - Special Session

Aim and Scope.

---

Hypercomplex-valued neural networks (HVNNs) constitute a growing research field that has attracted continued interest for the last decade. Complex and quaternion-valued neural networks are examples of hypercomplex-valued models, which also include tessarine and Clifford-valued neural networks. Complex-valued neural networks are essential for adequately treating angle and the information contained in phase, including the treatment of wave- and rotation-related phenomena such as electromagnetism, light waves, quantum waves, and oscillatory phenomena. Quaternion-valued neural networks, which have potential applications in three- and four-dimensional data modeling, have been effectively used to process and analyze multivariate images such as color and polarimetric SAR images. More generally, besides their natural ability to treat multidimensional data, hypercomplex-valued neural networks can benefit from the geometric and algebraic properties of hypercomplex algebras.

Despite significant theoretical development and successful applications, there are still many research directions in HVNNs, including a formal generalization of the commonly used real-valued network architectures and training algorithms to the hypercomplex-valued case. As powerful machine learning techniques, there are also many exciting applications of HVNNs for signal processing, including pattern recognition, nonlinear filtering, and prediction.

This special session welcomes papers that are or might be related to all aspects of hypercomplex-valued neural networks, including complex-valued and quaternion-valued neural networks. Papers on theoretical advances and contributions of applied nature are all appreciated. We also welcome interdisciplinary contributions from other areas on the borders of the proposed scope.

This special session aims to be an excellent forum for exchanging ideas on HVNNs for signal processing. We hope the proposed session will attract potential speakers and researchers interested in the interface between the theory and applications of HVNNs. We also expect this session to benefit and inspire researchers and practitioners that need sophisticated machine-learning tools for signal-processing applications.

Organizers.

---

We are looking forward to seeing you!