We propose semi-supervised regression and dimensionality reduction methods for hyperspectral subspace learning based on abundant unlabeled data and a small number of labeled data. The quantitative target variables for regression and the order constraints for dimensionality reduction are embedded in matrices representing data relations, i.e., a set of between-class scatter matrices, within-class scatter matrices and supervised local attraction matrices. The optimal projection matrices are estimated by generalized eigenvalue problems based on the matrices. The proposed methods are applied to dimensionality reduction problems based on a time-series of hyper-spectral data for a deciduous broad-leaved forest to extract local coordinates related to phenological changes.
