This package hosts a collection of CCA implementations in MATLAB and takes advantage of MATLAB's fast decomposition routines (SVD, eigendecomposition, QR, Cholesky, etc.).
See script_scale.m to run the experiments with timing and correlation coefficient comparisons.
Set the parameters of the random matrices through lines 3-6, and switch between sparse and dense matrices by commenting in lines 9-14 (for dense matrices) and commenting out lines 17-21 or vice versa (for sparse matrices).