I would like to separate the training and application processes so I could train method only once and then use it several times. Particularly, I need the information about number of training signal and background events at the application level. It seems I cannot get it from xml weights file. So is it possible to get that info somehow? (Probably using TMVA::Reader).
I tried:
The reader takes all its information from the xml-file, so unless it’s there it cannot be retrieved.
If this is related to your other question I suggested a sketch for a workaround grabbing the efficiencies from the ROC curve there.
If you are interested in the total raw number of sig and bkg events you could read these from your input data, e.g. TTrees which can be done, although a bit inconveniently, at application time.