-
remove explicit output variable scaling from svm transform/distance
-
Address some issues with classification and regression
-
When training transforms involving pipes, and projecting from one stage to the next, break the training data into single item templatelists (similarly to what is typically done by Distribute transforms durign training). This allows e.g. ExpandTransform to work as expected.
-
If 'subject' is set, but 'label' is not set, before remapping subject to an integer, check if it is already a number. In this case, no remapping is necesasry, and in fact remapping caused problems when evaluating regression output
