TMVA  checks (19/11/18)
----------------------


Two inputs (signal + background)
--------------------------------

 Fig 1 MLP shows the input data ( histxy) and 
 the selected components histxy0 and histxy1 for a neural net selection.

 Fig 2 BDT shows the input data ( histxy) and 
 the selected components histxy0b and histxy1b for  adaptive boost
 ( default constants).

 Fig 3 BDT shows the input data ( histxy) and
 the selected components histxy0b and histxy1b for  adaptive boost
 (constants times 3 ).


 Fig 4 BDTB bagging (default constants).


 Fig 5 BDTG gradient (default constants).


 Fig 6 BDTD Fisher discriminant for node
 splitting  (default constants).





Three inputs ( 2 as signal, 1 as background  and converse)
----------------------------------------------------------


 Fig 1 MLP shows the input data ( histxy),
 background (histxy0), signal(histxy0) and the NN response for
 inputs 1 and 2 as signal and input 3 as background.
 In contrast to Jetnet,  separation into three signals in one
 analysis does not appear possible.

 Fig 2 MLP as previous, but defining the input 2 as signal and
 inputs 1 and 3 as background. NN is 2 10 10 1 . Some miss-selection is evident.