Dynamic artificial neural networks based on the target feature and aplication in target recognition

The dynamic RBF artificial neural networks (ANNs) is put forward in the paper, which aims at only recognition of the target feature. It does not search the separating hyperplane of the whole space, but searches the separating hyperplane of the local space taking the target feature as center. To show better importance of each sample to the target feature, a method is researched that expected output of the dynamic ANNs training process is measured. And the dynamic training set is reconstructed and controlled dynamically according to the expected outpute. At last, the dynamic RBF ANNs is applied to the underwater acoustic target recognition that is utmost important to submarine war. Experiment results show that it is more robust than the traditional ANNs.