At present， about the definition of neural network are not unified， according to the neural network at Hecht， Nielsen's point of view， the neural network is defined as: "the neural network is made up of a number of very simple processing units to each other in some way connected to form a computer system， the system depends on its status， the dynamic response of the external input information to process " The sources and characteristics of integrated neural network and a variety of interpretation， it can be expressed as: simple neural network is a designed to mimic the human brain structure and function of information processing From the perspective of the theory model of neural network， it can be divided into two main categories; Namely layered feedforward neural network， neural network and Internet feedback for layered feedforward neural network， its characteristic is through appropriate BP algorithm for sample training and learning of the neural network input and output can approximate arbitrary input and output the corresponding nonlinear mapping， that after learning to implementing a map is a kind of adaptive control Due to the layered feedforward neural network has "self learning" and "training" function， and can imitate the intelligence of the human brain， thus has strong classification and Therefore， layered feedforward neural network is widely used in network communication channel equalization， the global network management， information flow prediction as well as other adaptive control， And the characteristics of the feedback neural network is through design to study the associative memory content or optimization answer set to minimum point of the system energy function， the dynamic equilibrium process of neural network can realize the automatic rapid processing optimization problems， so it can be widely used for the field of network communication， including the selection of packet scheduling and optimal routing， exchange of information and High speed Internet， nonlinear neural network also has the chaotic behavior， it is a very complex NP problem， can produce unpredictable sequence trajectory， fast password algorithm can be designed to be safe and It is because of the neural network learning mapping， lenovo optimization function and the characteristics of chaotic behavior can in theory to resolve the broadband network communication technology is facing some problems， which has been widely used in network 3 neural network application in network communicationThe following according to the "self learning" function of neural network， lenovo optimization function and chaotic behavior of the three functional features from three aspects: introduces the application of neural network in network communication instance求采纳
