Nonlinear Principal Component Analysis on Real-World Data

摘要

Principal Component Analysis (PCA) is an useful method in imiltivariate analysis to reduce a dimensionality of data. We have already proposed a nonlinearly extended model of PCA and have shown its effectiveness with some artificial data. In this paper, we report results of a nonlinear principal component analysis on real-world data utilizing the proposed method. Moreover, we compare the distribution of transformed data by nonlinear mappings with the distribution of the original data to discuss a nonlinearity of the data distribution.