A broad class of discrete-time hypercomplex-valued Hopfield neural networks.

Area of Science:

• * Computational Neuroscience
• * Algebra
• * Artificial Intelligence

Background:

• * Hopfield-type neural networks are crucial for associative memory and optimization problems.
• * Stability analysis is essential for guaranteeing reliable network performance.
• * Hypercomplex numbers offer a richer framework for neural network models beyond traditional real or complex numbers.

Purpose of the Study:

• * To introduce novel hypercomplex number systems and activation functions for discrete-time Hopfield-type neural networks.
• * To establish theoretical conditions for ensuring the stability of these networks.
• * To extend the stability analysis to a broader class of hypercomplex-valued neural networks, including those based on Cayley-Dickson algebras.

Main Methods:

• * Development of real-part associative hypercomplex number systems, generalizing existing algebraic structures.
• * Introduction of B-projection functions for hypercomplex-valued activation potentials.
• * Application of theoretical frameworks to confirm stability of existing networks and analyze new classes.

Main Results:

• * The proposed real-part associative hypercomplex number systems and B-projection functions ensure network stability.
• * Stability analysis is confirmed for several existing discrete-time hypercomplex-valued Hopfield-type neural networks.
• * A general class of Hopfield-type neural networks on Cayley-Dickson algebras is introduced with stability analysis.

Conclusions:

• * The novel hypercomplex number systems and activation functions provide a robust framework for stable neural network design.
• * The study enhances the theoretical understanding of hypercomplex neural networks, enabling broader applications.
• * This work lays the foundation for further research into advanced hypercomplex neural network architectures and their stability.