Abstract
Abstract
Synaptic memory depth plays an important role in neural synchronization and memory consolidation, while a non-volatility memristor provides an effective model to explore memory-dependent neuronal dynamics inspired by long-term synaptic plasticity. In this paper, a discrete memristor model with switchable volatile and non-volatile states is proposed and incorporated into heterogeneous neural networks to investigate the influence of synaptic memory characteristics on synchronization behaviors. Numerical results show that the critical coupling strength required for synchronization of the memristive-coupled neurons are significantly reduced as the memristor transits from volatile to non-volatile. The synchronizability of the memristive-coupled neurons is closely related to local activity of the non-volatile states of memristor. Furthermore, based on a local Lipschitz stability analysis, a synchronization criterion and a quantitative relationship between the critical coupling strength of the memristive-coupled neurons and the non-volatility parameter of the memristor are obtained, showing good agreement with the numerical results. Finally, FPGA-based hardware of the memristive-coupled neurons is setup to verify effects of the non-volatility of memristor on the synchronizability of coupled neurons. This work provides a nonlinear dynamical perspective for understanding the role of long-term synaptic plasticity in neural coordination and memory-related collective behaviors.