Improvement of neural network learning algorithm and its application in control
Wu Yan, Hongbao Shi · 2002
With the drawbacks in learning algorithm of neural network taken into consideration, fuzzy logic is integrated into the neural network and its learning process to improve its performance. A fuzzy neural network named GFNN, its corresponding off-line learning algorithm, and on off-line learning algorithm named fuzzy backpropagation (F-BP) are proposed in the paper. These learning algorithms greatly speed up the learning process of neural netwporks. In addition, online learning algorithms of F-BP and GFNN are also proposed so that these neural networks can adapt dynamically to the environment by revising the parameters of the neural networks. To prove their effectiveness, the proposed neural networks and learning algorithms are used to simulate the train operation control system, which has produced a very good test result.