Fuzzy Neural Network Research And Its Application In Pattern Recognition

In the numerous artificial intelligence research areas, a mass of researches focused on what many kinds of intelligence recognition methods inosculated, in order to improve the effect of the recognition. Following the artificial intelligence technology development, more and more intelligent recognition algorithms were proposed in the pattern recognition domain. The fuzzy neural network combined fuzzy logic which has an ability of processing indefinite information with neural network which has the knowledge memory property, displayed their respective superiority, and makes up respective insufficiency, including learning, association, recognition, auto-adapted and fuzzy information processing as the whole, improved system's learning capability and the power of expression. The fuzzy reasoning process with the neural network can gain the rule from the input /output data using the network study or the cluster method, make the auto-adapted adjustment to the rule under the performance index instruction, and realize auto-study, auto-adaptation of the fuzzy system. It has the unique superiority in the pattern recognition and the classification, and attracted the eyeballs of many experts in this domain when we used this technology into the domain of pattern recognition. That is a hotspot in the current research.The fuzzy neural network based on the neural network, in view of this characteristic, put forwarded an efficient algorithm which can solve transportation problem on the basis of unified effectively the Hopfield neural network and the Adams formula. Combined the optimization function of the Hopfield neural network with the actual situations and transferring the issue of the optimization of transportation to the break-even point of the network system, this algorithm can solve commendably the problem of transportation. Meanwhile, according to multi-step thought of Adams propose a method which based on neural network parameter adjustment of many momentum revisions, that algorithm is suitable for many kinds of forward neural networks. The experiment indicated that this plan can be characterized on avoiding the risk of vibration and quick convergence after setting up various different initial parameters.Putting forward a new fuzzy neural network model combined with various different NN models, which is benefit to not only realized the synthesis of the input variable's membership function value in second layer, but also matched automatically the fuzzy control rule, moreover, it can not need to consider expert provide the fuzzy reasoning rule for fuzzy neural network, and nodal point number are few in the inference layer, which not only simplified the complexity of network so that the algorithm is much easier to calculate and program, but also realized the recognition and forecast of earthquake magnitude well according to applying the fuzzy neural network model into the specific pattern recognition. It has a good value on utility.