The relationship between input data and output data in TMVA

TMVA
1

I find that several options for reading input data in TMVAGUI are (1b)'Deco' -transform and (1c)'PCA' -transform and (3)Linear Correlation, but I don’t quite understand that (3)Linear Correlation here is based on initial data. Or based on data that has already been transformed？

And the(4b) final classification result, is it based on the initial data or the data that has been transformed?

image
image423×653 39.3 KB

If both are based only on the initial input data, what is the role of these two transformations here?

Hope to have enthusiastic personage can answer my question, thank you very much!

Cheers,
Kevin.

2

Hi @chaerme; I am sure @moneta can reply to your question.

Cheers,
J.

3

Hi,

These transformations are performed on the input data before being used by the TMVA method. If you are using transformations, the final classifier results will be based on the transformed data.

Lorenzo

4

Hi, @moneta
I don’t quite understand what you mean.

TMVA::Factory* factory = new TMVA::Factory(“TMVAtest”, outputFile,

  "!V:!Silent:Color:DrawProgressBar:Transformations=I;D;P:AnalysisType=Classification");

Does that mean that as long as I select the transformation here, everything that follows will be based on that transformation? And when I use the classification method later, don’t I need to set the transformation option?

5

Hi,
If you put the Transformation in the Factory they will be applied to the methods, they will just be tested and produce some information plots. To have the transformations used for the classification by the method you need to specify them in the BookMethod. See paragraph 4 (page 48) of the TMVA Users Guide

Lorenzo

1 Like
6

@moneta

I get it ! Thank you !

Cheers,
Kevin