The pylink model will return a smaller subset of training samples, if some model runs were broken off. These should then be used for the training.

The analytical example runs with validation errors in the scale of 1 if 'is' is used. This reduces to validation error e-7 for '=='

Preprocessing includes building the InputSpace, checking the input dimensions and assigning the inputs to parameters.

Some surrogate strategies and PostProcessing options need to be reworked

Tested static training with PCE on analytical example, remaining changes tbd for sequential training and postprocessing

Initial split of the classes, MetaModel contains empty or general functions, PCE adds PCE/aPC specific options.
Training tested with User-guide example