|
Neural networks because of its unique distributed information storage and self-organized self - learning ability , as well as parallel processing , superior performance , scholars in the field of decades of unremitting efforts , has been in the optimization calculation , fault diagnosis , associative memory , pattern recognition , etc. has demonstrated good ability , and has been widely used in the actual field of industrial control , clinical , commercial bank loan risk warning . However, these far not fully reflect the advantages of the neural network , to make the neural network to play a more powerful application potential , first of all deal with the complexity of the network structure analysis , familiar with its dynamic characteristics , including stability, passivity . Firstly passive analysis of cellular neural networks with time-varying delay , and then consider the linear fractional parameter uncertainties , it is determined that the system global robust passive full guidelines . Compared with relevant literature , this paper introduces the Delay segmentation method effectively reduces the conservative nature of the results ; then discussed passive cellular neural networks with distributed delays and infinite delay , suitable inequality zoom , by introduce free weight matrix , the relaxation of the restrictions on the conditions of the time delay function . The final analysis of a class contain some time varying delay and neutral delay neural network model , finite time delay system under passive general conclusions can be directly applied to single- varying delay system . Lyapunov-Krasovskii stability determination method research methods based primarily on structure containing information Delays split Lyapunov functional , introducing more slack variables , and considering the usually ignored or exaggerate items , and take appropriate inequality transform relatively better passive criterion . In addition , the selected network model in this paper in various time delay function, and the neurons connected functions, and so does not limit into identical , making the results more general . Finally combined Shu complement theorem , the passive criterion of the robustness of the system in the form of linear matrix inequalities . Conclusion of each chapter are made ??to the number of cases of simulation using MATLAB toolbox , you can easily verify the feasibility of the conclusion .
|