Karhunen Loeve Mapping, followed by scaling
[W,FRAC] = KLMS(A,N)
First a Karhunen Loeve Mapping is performed (i.e. PCA or MCA on the average prior-weighted class covariance matrix). The result is scaled by the mean class standard deviations. For N and FRAC, see KLM.
Default N: select all ('pre-whiten' the average covariance matrix, i.e. orthogonalize and scale). The resulting mapping has a unit average covariance matrix.