Fuzzy Neural Network Research Summary

Along with science and technology rapid development, the people also were unable to avoid fuzziness again, this needed to have the processing fuzzy method. In view of this question, in this article, I has carried on the summary to the fuzzy neural network.This article based on the fuzzy mathematics and the neural network, introduced the fuzzy neural network theory by and its the domestic and foreign research present situation. From the fuzzy neuron, the fuzzy neural network type, the fuzzy neural network modular design and the study algorithm several aspects have carried on the introduction to the fuzzy neural network model. Finally, has outlined the fuzzy neural network in the complex industry modelling and the control domain application.The first part has carried on the summary to the fuzzy neural network survey, introduced to at present already some commonly used methods which has, including the fuzzy pattern recognition, gathers a kind of analysis, the information retrieval, the similar choice, the synthesis judgment, the fuzzy control, the probability analysis and the word card analysis and so on.The artificial neural network is, with simulates the human brain behavior which becomes by the massive simple parts widespread connection the complex network system. The artificial neural network is to the biology nervous system simulation, its essence is the non-linear mathematics theory application.The artificial neural network has the parallel information-handling capacity, the distribution information storage ability, as well as displays in the study process from the organization sense, auto-adapted and the fault-tolerant characteristic.The artificial neural network has received the characteristic which in the biology nervous system the information processing and the information storage Fusion. In the traditional computer, the computation and the memory are the completely independent two parts. The artificial neural network technology from essentially has solved this problem, because its movement is from inputs to the output, between each artificial neuron pitch point the value transmission process, has also completed the information memory and the computation while value transmission.The neural network has the very strong fault-tolerant ability. Because the information is not saves the memory in the single neuron, but is the distribution memory in the entire neural network, and manifests between the massive neurons in the synapse union intensity. Like this when the few neurons receive the damage, through from the organization function, the overall pattern which other multitudinous neurons will compose causes the network the overall function to continue in force.Will blur the technology and the neural network carries on the organic union will be fuzzy technical and the neural network thorough research and the development one kind of inevitable trend.Along with fuzzy information processing technology and neural network engineering research unceasingly thorough, will blur technical and the neural network technology carries on the organic synthesis, thus the structure will leave one kind to be possible "to be automatic" the processing fuzzy information fuzzy neural network or called the auto-adapted fuzzy system, has aroused the more and more many scientific and technical worker's research interest and the attention, will become the current important research hot spot.Fuzzy technical and the neural network technology fusion may effectively display respective superiority and make up respective insufficiency. Uses the fuzzy technology to be easy to carry on Gao Jie the information processing, will blur the technical introduction neural network, might greatly open up neural network processing the scope and the ability, will enable its not only to process the precise information, also will be able to process the fuzzy information or other not accurate information, not only will be able to realize the accurate association and the mapping, but also might realize the inaccurate association and the mapping, specially fuzzy association and fuzzy mapping. The neural network has the very strong superiority in the study and the automatic pattern recognition aspect, uses the neural network technology to carry on the fuzzy information processing, then causes the fuzzy rule automatically to withdraw and the subordination function automatic production has the possibility to be able to solve, causes the fuzzy system to become one kind of auto-adapted fuzzy system.The fuzzy method and the neural network two is supplementary. They all stress to processing are actual the systems control question which by the determinism, the inaccuracy does not cause. But the difference lies in the neural network quite to be difficult regarding knowledge extracting with the expression, the fuzzy information processing method actually is good at simulating person's experience to process some indefinite information, Moreover, fuzzy system very difficultly from sample direct study rule, also can increase fuzziness in the fuzzy inference, but the neural network has the stronger learning capability, may reduce fuzziness using the association memory. Has like this had the fuzzy neural network forecast method, it is in the modern soft computation concept important content.The second part summarized the fuzzy neural network model, the fuzzy neuron study and the auto-adapted mechanism are, body which revises through the power value and the body realizes revises to the neuron structure carries on the revision. The synapse revises including takes the neuron weight the mapping function and the subordination function change; The body revises including in the rule change, the rule the fuzzy variable subordination function change as well as the rule expression way change.The fuzzy neural network (Fuzzy Neural Network, FNN) is may carry on the fuzzy information processing the neural network. They usually are a kind by massive are fuzzy or the non- fuzzy neuron mutually join the constitution the network system. The fuzzy neural network has the many kinds of types, is similar with the general neural network front, usually, divides into it to the fuzzy neural network and the feedback fuzzy neural network two big kinds.The neural network had the non-linear mapping ability, the learning capability, the parallel processing ability and the fault-tolerant ability, the fuzzy logic had the processing not definite ability, the two already obtained the widespread application in the complex industry object modelling and the control domain. Because the neural network itself question (for example exudes ability not to be able to achieve control system robust request) beside, the neural network black box -like knowledge expression way enables it to use the known Knowledge carries on the study.The pure use neural network control technology research has the tendency which bogs down. But the fuzzy logic application has also met the question which the fuzzy rule difficulty with determined. The fuzzy neural network unified the fuzzy logic and the neural network merit, has avoided the two shortcoming, both may have the fuzzy logic the indefinite information-handling capacity, and may brightly after the network from the learning capability, therefore has the very widespread application prospect in the control domain.The fuzzy theory and the artificial neural network have the logical inference and the value computation function and the stronger nonlinear function approximate ability, moreover they all do not need to establish the precise mathematical model, may process not the precise even information realization smooth inference, the fuzzy neural network is collects the two the merit, the collection training, the association, auto-adapted and the fuzzy information processing to a body one kind of special artificial neural network.The third part has carried on the summary to the neural network application, based on the commercial bank credit risk assessment progress and the limitation, appraises the commercial bank credit risk with the fuzzy neural network. The system mathematical model in the controller design process and in the control system performance analysis, is playing the indispensable role. Also may apply to the geological disaster studies, the water quality appraisal, the weather forecast and the synoptic situation analysis, the industry and the transportation system, the artificial intelligence, the medical diagnosis, the commodity judgment and so on.The fuzzy neural network theory and its application have obtained the rapid development, each kind of new fuzzy neural network model proposed and the study algorithm research which if adapted not only has accelerated the fuzzy neural theory consummation, moreover also obtained the extremely widespread application in the practice. Although fuzzy neural network controlled city commander-in-chief in the complex system application and so on in modelling has already obtained the very big success, but took a new technology, it still had some problems in the theory and the application.The fuzzy neural network has gathered the neural network and the fuzzy control merit, becomes the intelligent control research the hot spot, its research basic content mainly includes the model structure the determination, the study algorithm research, the stable analysis, the controller design and the auto-adapted fuzzy neural network design and so on.In brief, the fuzzy neural network gradually the network architecture which will approach to the intellectualization, unceasingly will enhance the network the computation ability, will strengthen the network the dynamic handling ability, and will cause the network the structure as far as possible simple also is easy to realize, fuzzy neural network structure normalization, strengthened different application domain versatility.