| Summary: | 碩士 === 國立彰化師範大學 === 工業教育與技術學系 === 97 === This thesis presents the complete study of stability analysis and state estimator design. The system is focused on discrete stochastic recurrent neural networks with interval time-varying delays. For a stability analysis problem, the goal is to develop globally robust delay-dependent stability for uncertain discrete stochastic recurrent neural networks with interval time-varying delays. The parameter uncertainties are assumed to be time-varying norm-bounded in the neuron state and activation functions. The activation functions are assumed to be globally Lipschitz continuous. Based on the Lyapunov-Krasovskii functions combining with linear matrix inequality (LMI) techniques, globally robust delay-dependent stability criterion which is dependent on both the lower and upper bounds of the interval time-varying delays is derived by introducing some free weighting matrices. For the estimator design problem, state estimation for discrete stochastic recurrent neural network with interval time varying delays was investigated. Attention was focused on the design of a state estimator which ensures the global stability of the estimation error dynamics. A delay-dependent condition on both the lower and upper bounds of the interval time varying delays is given in terms of a linear matrix inequality to solve the neuron state estimation problem. When this LMI is feasible, the expression of a desired state estimator is also presented. Finally, some numerical examples are provided to demonstrate the applicability of the proposed approach.
|
|---|
