Flag manifold Feature selection by matrix factorization

‎Feature selection is a challenging task in data analysis process‎. We constructed a flag manifold of isotropy subgroup, H, as the eigenvalues of features matrix from the high-dimensional data set X, then demonstrate its upper triangular matrix of these flags and the quotient space of GL /H‎. Applying the al‎‎gorithm, Regularized Matrix Factorization (RMFFS) for unsupervised feature selection, on this quotient-space instead of X will reduce dimension of the dataset X and we call this new algorithm FMFS algorithm‎.