The dynamics of a neural network under several factors (bias current and electromagnetic induction effect) are recently used to simulate activities of the brain under different excitation. In this paper, we introduce a novel Hopfield neural network (HNN) based on two neurons with a memristive synaptic weight connected between neuron one and two based of flux controlled memristor recently proposed by Hua M. et al., in 2022. Using analysis tools, we proved that this model can develop rich dynamical characteristics such as various number of equilibrium points when the parameters are varied, four-scroll attractors, transient chaos, multistability of more than three different attractors and intermittency chaos phenomenon are reported. Moreover, when increasing a synaptic weight, the model shows bursting oscillations phenomenon. To obtain the normal state of the brain, the control of multistability to a strange monostable state is carry out. Finally, microcontroller implementation of the model is considered to verify the numerical analysis.
This work deals with the regular and chaotic dynamics of a system made up of two Hopfield-type neurons with two different activation functions: the hyperbolic tangent function and the Crespi function. The mathematical model is in the form of an autonomous differential system of order four with odd symmetry. The analysis highlights nine equilibrium points and four of these points experience a Hopf bifurcation at the same critical value of a control parameter which can be either the diss1ipation parameter or one of the coupling coefficients. This makes plausible the presence of four parallel bifurcation branches as well as the coexistence of multiple attractors in the behavior of the system. One of the highlights revealed in this work is the coexistence of three double-scroll type attractors of particular topology as well as the presence of a four-spiral attractor. Furthermore, the coexistence of both self-excited and hidden dynamics is also reported. All this plethora of dynamics is elucidated by making use of the usual tools for analyzing nonlinear systems such as bifurcation diagrams, the maximum of Lyapunov exponent, basins of attractions as well as phase portraits. A physical implementation of the microcontroller-based system is envisaged in order to confirm the plethora of behaviors observed theoretically.
16 August 2024 in The European Physical Journal Special Topics
This work studies the dynamics of a three dimensional Hopfield neural network focusing on the impact of bias terms. In the presence of bias terms, the models displays an odd symmetry and experiences typical behaviors including period doubling, spontaneous symmetry breaking, merging crisis, bursting oscillation, coexisting attractors and coexisting period-doubling reversals as well. Multistability control is investigated by employing the linear augmentation feedback strategy. We numerically prove that the multistable neural system can be adjusted to experience only a single attractor behavior when the coupling coefficient is gradually monitored. Experimental results from a microcontroller based realization of the underlined neural system are consistent with the theoretical analysis.
27 April 2022 in Physica A: Statistical Mechanics and its Applications