Linear classifier using PC expansion on the joint data.
W = PCLDC(A,N)
Finds the linear discriminant function W for the dataset A computing the LDC on a projection of the data on the first N eigenvectors of the total dataset (Principle Component Analysis).
When ALF is supplied the number of eigenvalues is chosen such that at least a part ALF of the total variance is explained.
If N (ALF) is NaN it is optimised by REGOPTC.