【摘要】 为了研究神经网络在复杂电磁环境中的动力学行为，提出了一种电磁辐射下的分数阶忆阻耦合Hopfield神经网络模型.基于余弦函数和Sigmoid函数序列设计一种具有任意多稳态和强记忆特性的新型分数阶局部有源忆阻器，用于模拟神经网络模型的电磁辐射效应.分析该模型的非线性动力学行为，主要包括相图、平衡点、分岔图、谱熵复杂度.结果表明，该模型随阶次、参数、初值变化时，具有对称的共存周期吸引子、对称的共存多涡卷吸引子、单方向多涡卷吸引子、共存吸引子等复杂动力学行为.此外，该模型具有谱熵复杂度为0.435 1的多涡卷隐藏吸引子，能更好地应用于信息安全领域.更多还原

【Abstract】 To study the dynamic behavior of neural networks in complex electromagnetic environment, a fractional-order memristor coupled Hopfield neural network model under electromagnetic radiation is proposed.A novel fractional-order local active memristor, designed using cosine and Sigmoid function sequences, characterized by arbitrary multi-stable and strong memory effects, is introduced to simulate the electromagnetic radiation effect of the neural network model.The nonlinear dynamic behavior of the model is analyzed, with phase diagrams, equilibrium points, bifurcation diagrams, and spectral entropy complexity.The results show that the model exhibits complex dynamic behaviors, such as symmetrical coexistence periodic attractors, symmetrical coexistence multi-scroll attractors, unidirectional multi-scroll attractors and coexistence attractors, influenced by changes in the orders, parameters and initial values.In addition, the model has a multi-scroll hidden attractor with a spectral entropy complexity of 0.435 1, which can be better applied to the field of information security.更多还原