Algorithms Of Training Multilayer Neural Networks In The Automatic Control System
Annotatsiya
The application of neural network technologies in control systems consists of several stages. In technological object control systems, each layer of the multilayer perceptron neural networks is represented using separate transfer functions. The result of synthesizing neural network technology into control systems allows for the rapid solution of many problems. Using neural network technology, it is possible to perform many calculations. Chain rules of neural networks are added this paper. Lots of training algorithms are compared to choose better to train network. Backpropagation neural networks is seen deeply in the last page. Problems of interpolation and extrapolation are solved. Algorithms for constructing an approximating function based on a multilayer perception neural network are seen. Training stages and algorithms of multilayer neural networks for the system are concluded. Interpolation and extrapolation problems in neural network technology are set there.