This reverts a change we had made in the previous version:
When input is scaled to correlations (i.e. do_scaling=True), then the solution/precision matrix and low rank component are rescaled to covariances. The optimal regularization parameters however are not rescaled!
This reverts a change we had made in the previous version:
When input is scaled to correlations (i.e.
do_scaling=True), then the solution/precision matrix and low rank component are rescaled to covariances. The optimal regularization parameters however are not rescaled!