Nonlinear principal component analysis based on RBF neural network and principal curve

作者：Jia Mingxing^*; Zhao Chunhui; Wang Fuli; Mao Zhizhong; Li Hongru

来源：Chinese Journal of Scientific Instrument, 2008, 29(3): 453-457.

On the basis of analyzing non-linear principal curve, a non-linear PCA method based on clustering, linear principal component analysis and neural network is presented in this paper. The proposed method has the following advantages over traditional methods: first, it has conceptually the same explanation as linear PCA, and gives the calculation methods of nonlinear principal scores and loadings; second, it adopts three-layer neural network, which is simple in structure and easy to be trained. The data samples used for training neural network are obtained by clustering and linear PCA method, which solves the problem of training data insufficiency. The results of both simulation and three-tank experiment demonstrate the effectiveness of the proposed method.

• 出版日期2008