Analysis And Improvement Applications Of Bp Neural Network

With the rapid development of science and technology, so does artificial neural network. And it is also important in the intelligence field. It has already extensively applied to many fields, such as handle singles, mode identify, machine control, expert system. And it has got more thorough development in predicting field. BP neural network is a typical algorithm in the artificial intelligence network, and it has very strong nonlinear mapping capability. Its most outstanding ring is to solve some nonlinear problems. Also it has many other advantages, its network topology structure is simple, it has high margin precision, it is easy to program, and it has very strong maneuverability, etc. Therefore, the application of BP neural network becomes more extensive, and it has become one of the most important algorithms in intelligence field.Although Standard BP neural network has many advantages, also it has many disadvantages. Its optimization algorithm is based on the most soon calculate way algorithm. The most soon calculate way algorithm has many disadvantages, such as slow convergence speed, easily run into local minimum dot. Therefore, in the side the standard BP calculate way has the same disadvantages, and it leads to the shortage of study capability of network. In addition to the disadvantages due to the most soon calculate way algorithm, also it is hard to get the needing layers and the nodes in each layer of the implicit layer of BP network. And it is a kind of neural network with tutor. As the instruction information, each input mode must know the expect output and margin precision. To the disadvantages of standard BP network, many domestic and international scholars have done detailed research with that, and they put forward many enhanced algorithms.The paper aims at the disadvantages of BP algorithms, adds the basic principle of basic BP neural network, it puts forward an enhanced algorithm. The algorithm is based on the additional momentum method, adds a self adapt factor about study speed, and carries on deduction of that. Through applying the new algorithm to solve the XOR problem, it can be used to validate the possibility and robustness of the new algorithm. Through the research, we know that the new algorithm has strong stability and robustness.