Deal all TMVA experts,
I have seen the two visualized functions of Control Plot and Decision Boundary in BDT, but i have searched a lot, seems there is not that many materials related to them, even in the TMVA user guide.
Anyone knows how to interpret the Control Plot(I guess it’s similar to “Parallel Coordinates” in the GUI? But for “Parallel Coordinates” there are so many different color lines I am neither not sure of their meanings… ) to help improve the BDT and also how to get the Decision Boundary plot?
Where have you seen these plots?
Parallel coordinates is a way of visualising high-dimensional spaces and makes certain kinds of structure easily identifiable. In TMVA this is provided to visualise correlations between input variables and also output variables.
The control plot gives you information about the boosting process (weights and errors for the different iterations).
The decision boundary plot can be generated with the file
$ROOTSRC/tmva/tmvagui/src/mvaweights.cxx where $ROOTSRC is the path to the source code of ROOT.
Oh I just saw them in the user guide and the TMVA GUI, then tried to interpret them to help us understand what’s happening in the training. But it seems not that easy…
Thanks for the explanation.
And BTW I have another question on the Kolmogorov-Smirnov test(to estimate the overfitting) and hope you can help me with the print out:
Testing efficiency compared to training efficiency (overtraining check)
— Perform Kolmogorov-Smirnov tests
— Goodness of signal (background) consistency: 0.572 (0.047)
I have seen in the slides:
Does it mean that the ideal value of the Goodness of signal (background) consistency should be 0.5 for both signal (background) ?
Yes, in principle. You can also check the output classifier distributions with both test and training samples. With this plot you can visually inspect the similarity of the distributions.
Got it. Many thanks for the guidance!
Actually I have another quick question. I have searched on my local Root directory, it seems that there is only header file.
Do you know where we can find the detailed source file of all TMVA methods locally in the ROOT package?(like below, so that it’s convenient to see how the methods are written and then how to optimize them)
You can check out the git repository code at http://github.com/root-project/root/.
git clone http://github.com/root-project/root/ to download it easily
Many thanks for the help!