Comparisons between minimum error entropy and minimum mean squared error based on principal component neural networks

Principal component neural networks (PCNNs) give an adaptive parallel method to extract the principal components (PCs) and the principal subspace. But for non-gaussian stochastic system, PCNNs with minimum reconstruction squared error do not contain maximum information about original system definitely. In this paper, a PCNN based on minimum reconstruction squared error and its reconstruction property on squared error and entropy are introduced firstly. And then a PCNN based on minimum error entropy and a approximation method for the entropy are proposed. Finally, the computing results of the two indices are compared.