A Grassmann-Rayleigh Quotient Iteration for Dimensionality Reduction in ICA

摘要

We derive a Grassmann-Rayleigh Quotient Iteration for the computation of the best rank-(R_1, R_2, R_3) approximation of higher-order tensors. We present some variants that allow for a very efficient estimation of the signal subspace in ICA schemes without prewhitening.