Generation of a very large dataset
[A,U] = GENDATV(M,C,K,R,T)
This routine generates an arbitrarily large dataset of C classes by first generating the class means from a standard K-dimensional normal distribution. The class priors are genetated from a N(1,S) distribution, After that their sum is normalized to one. The classes themselves have spherical normal distributions with an equal standard deviation R for all features and all classes.
In U the class means, the class priors and R are returned, such that similar data from the same distribution can be generated.