几类时滞神经网络模型的全局稳定性-应用数学专业论文.docxVIP

摘 摘 要 本文主要研究了几娄时滞神经网络模型的全局稳定性(包括全局渐近稳定性。全局指数 稳定性和全局指数鲁棒稳定性)． 在第一章中，利用Lyapunov方法，线性矩阵不等式(LMI)方法和Halanay不等式技巧 研究了一类具有时滞的Cohen—O rossberg神经网络的全局渐近稳定性和全局指数稳定性。 在第二章第一节中，利用压缩映像原理、同胚性质和Lyapunov方法研究了一类定常时 滞神经网络的全局指数鲁棒稳定性，并得到了～些新的在工程上易于验证的判据；在第二 节中，利用非奇异M一矩阵理论、同胚性质和Halanay不等式技巧研究了一类具有变时滞 的神经网络的全局指数鲁棒稳定性。 本文所得到的结果是以前一些结果的推广相完善，并且本文的结果在工程上易于验证 和应用。另外，本文只要求激活函数满足全局Lipschitz条件，有时也假定激活函数是有界 的，但是不要求激活函数是可导的、单凋非减的，也不要求连接权矩阵是对称的，因此本文 的结果对神经网络的设计和应用具有重要的意义。 关键字，神经网络，时滞，线性短阵不等式，Halanay不等式，全局{{|i近稳定性，全局 指数稳定性，全局指数鲁棒稳定性。 Abstractin Abstract in this paper，we study the globm stability(the global asymptotica]stability，the global exponential stability and the global exponential robust stability)for several kinds of delayed rleural networks． In Chapter 1，by using Lyapunov method，linear matrix inequality(LMll approach and Halanay ineqality technique；we investigate the gobM asymptotical stability and the global exponential stability for a class of Cohen·Grossberg neural networks with delays In Section 1 of Chapter 2，by applying contraction mapping principle，homeomorphism property．and Lyapunov method．the global exponential robImt stability is discussed for a class of neural networks with constant delays．and a set of new criteria are obtained；In Section 2， by utilizing nonsingular M—matrix theory，hnmeomorphism property and Halanay inequMicy techniquej tim global exponential robust stability are considered for a class of neural networks with time—varying delays， The results obtained in this paper improve and extend some existing ones，and the given conditions are easy to verify and apply in engineering．In addition，in this paper，we only require the activation functions to be global Lipschitz continuous or bounded．not require the activation functions to be differentiable and nlonotone nondecreasing，and not require the in· terconnection matrices to be symmetry either Therefore the proposed criteria in this paper possess important leading significance in the design and application of neural networks