Change normalisation of PCA
tnm' = f(n) * tnm with f(n) = sqrt(n(n+1)/(2n+1)) and do then the PCA on the tnm'.
=> Diagonal operator 1/f(n) to take into account when doing the inverse transform for analyses (S matrix to modify).
tnm' = f(n) * tnm with f(n) = sqrt(n(n+1)/(2n+1)) and do then the PCA on the tnm'.
=> Diagonal operator 1/f(n) to take into account when doing the inverse transform for analyses (S matrix to modify).