The natural orthogonal composition (NOC) (Jackson et al. 1991;
Golovkov et al. 1992) is a technique that can be used for feature
extraction, i.e., given a set of multivariate measurements NOC
is expected to be able to provide a smaller set of uncorrelated
variables or components whose proper combination gives back
the larger starting set. In practice, given a set of observations,
it is possible to estimate a set of independent eigenvectors and
eigenvalues whose combination allows rewriting the observed
variables. Since this operation corresponds to transforming the
natural basis on which we observe the variable, into a new orthogonal
basis by diagonalizing the variance matrix, the NOC
technique consists essentially of a rotation of the old basis into
the new one.