Intelligent identification for discrete memristive neuron map: An adaptive chaos game optimization algorithm studied from the perspectives of different sample sizes and objective functions

Abstract Discrete memristive neuron systems have attracted considerable attention due to their nonlinear dynamical properties, low computational overhead, and ease of hardware implementation. For the practical engineering applications of discrete memristive neuron systems, effective control remains a key issue. Parameter identification using intelligent optimization algorithms is an important approach for controlling complex nonlinear systems. However, classical algorithms are prone to falling into local optima and often exhibit high computational complexity, resulting in slow convergence. Therefore, a new algorithm named adaptive chaos game optimization (ACGO) is proposed to address these issues. By introducing a differential evolution mutation strategy and a Cauchy adaptive parameter mechanism, the ACGO algorithm can effectively balance global exploration and local exploitation capabilities. To verify the effectiveness of the proposed algorithm, it is applied to parameter identification in five discrete memristive neuron maps (DMNMs) and compared with seven intelligent optimization algorithms. Simulation results demonstrate that the ACGO algorithm achieves higher accuracy and faster convergence. In addition, an in-depth investigation is conducted into the effects of sample size and objective function on identification performance. The results indicate that setting the sample size to 4 and selecting the mean squared error (MSE) as the objective function can achieve better identification performance and a high level of robustness.