Fixed-time stabilization of fully quaternion-valued delayed neural networks with discontinuous activations: An...

This paper presents a direct method based on exponential control for fixed-time stabilization of fully quaternion-valued neural networks featuring discontinuous activations and generalized delays. Firstly, a novel exponential control scheme is proposed and applied directly to the addressed quaternion-valued system, departing from conventional power-law control designs for real-valued subsystems obtained by separation. In convergence analysis, a non-decomposition-based direct analysis method is developed within the quaternion-valued framework to simplify the theoretical derivation and yield to concise stabilization conditions. Furthermore, Taylor expansion with full retention is employed for the exponential control function to reduce the estimate conservatism in the stabilization time. Finally, numerical examples are provided to validate the proposed method and theoretical findings.