Increased-Order Generalized Synchronization in Memristive Neurons: Analytical and Data-Driven Approaches

Abstract

We investigate increased-order generalized synchronization between four- and five-dimensional memristive Hindmarsh--Rose neurons, focusing on the selection and stabilization of the synchronization map. An analytical construction combines a prescribed nonlinear map with adaptive control, retaining the memristive switching contribution explicitly in the error dynamics. Lyapunov analysis establishes asymptotic synchronization under uniform strict dissipativity and yields a switching-independent decrease estimate outside an explicit error neighborhood. Numerical simulations exhibit error decay beyond the sufficient convergence conditions. A complementary data-driven construction uses next-generation reservoir computing to learn predictors from uncoupled full-state trajectories and an affine synchronization map through dynamical consistency. The resulting sampled-data controller achieves small residual tracking errors without access to the governing equations. Relative to a reference map within the data-driven framework, the learned map reduces median control effort approximately eightfold while improving tracking accuracy. These results connect synchronization-map selection with conditional convergence guarantees and reduced control requirements.