i’m using tmva 4.0.7 to do rectangular cut optimization as
in root/tmva/test/TMVAClassification.C. i’m experiencing
some strange behavior with some subset of the training sets
i’m using, and this behavior is reproducible. although the
background and signal efficiencies as well as the cuts generated
from the training set (and reported in the output) are sensible,
the final roc curve generated from the test sample is trivial, i.e.
B=S everywhere. this only occurs for some of the data sets i’m
using, and i wonder what’s going on and how i might be able to fix it.