A novel chaotic neural networks and application
Shi Wei-feng, Shilong Xue · 2005
To increase ability of modeling and identification for nonlinear system by neural networks, the characteristics of neurons, learning rule and configuration of networks are researched. The chaotic neuron is introduced to neural networks to form local recurrent chaotic neural networks. The information treatment quantity of the networks is all so increased because there are feed back loops with recurrent networks. The local recurrent chaotic neural networks are used for a marine synchronous generator modeling with a marine real time simulator. In the networks training of generator modeling, a dynamic BP learning algorithm is applied. Compare to other neural networks modeling, the neuron number of hidden layer of the networks is few, the ability of generalization of the chaotic networks system is well.