Chaotic Dynamics of Memristor-Coupled Tabu Learning Neuronal Network
Quanli Deng, Chunhua Wang, Gang Yang
Abstract
This paper presents an investigation into the dynamics of a Memristor-Coupled Heterogeneous Tabu Learning Neuronal Network (MCTLNN) comprising distinct types of Tabu learning neurons. The network adopts both monotone and nonmonotone composite hyperbolic tangent functions as activation mechanisms. A sinusoidal function-based memristor interconnects the neurons, leading to intriguing dynamical interactions. A theoretical analysis reveals the existence of infinitely many equilibria in the MCTLNN. Utilizing multiple numerical simulation methods, including phase portraits, Lyapunov exponents, bifurcation diagrams, and local attraction basin analyses, we systematically investigate the system’s complex dynamics. This study discovers the coexistence of diverse attractors, such as heterogeneous, symmetric, and initial offset-boosted attractors. Notably, we explore the dependence of the system’s dynamics on the initial conditions, specifically highlighting the initially offset boosting phenomenon through extensive simulations. Analog circuit-based hardware implementation not only confirms the model’s theoretical predictions, but also provides a platform for brain-inspired computing applications.