TTreeReader Error when applying MVA

ROOT
,
1

Hello experts,

I’m trying to now apply and save the results of my trained MVA to my ntuples and am currently trying to test this with the current training file. I seem to be running into some sort of issue due to TTreeReaderValueBase::CreateProxy() not being able to read a specific branch which is formatted as a double. Is there a solution to this? I’ve attached the code for both the MVA and application and there is a link to the example ntuple in google drive.

Edit: The code in question which causes the error is the Belle_e_tmva_apply.C

Error in <TTreeReaderValueBase::CreateProxy()>: The branch weMissM2__boY4ScleanMask__cm__sp0__bc contains data of type double. It cannot be accessed by a TTreeReaderValue<float>
terminate called after throwing an instance of 'std::runtime_error'
  what():  An error was encountered while processing the data. TTreeReader status code is: 6

ROOT Version: 6.32.00
Platform: Linux Red Hat Enterprise Linux 9 v9.6
Compiler: g++ (GCC) 14.2.0

Belle_e_tmva_apply.C (700 Bytes)
Bzptn_e_belle_tmva.C (4.8 KB)

2

Hi,
in your Bzptn_e_belle_tmva.C you have

    dataloader->AddVariable("weMissM2__boY4ScleanMask__cm__sp0__bc", 'F');		// Missing Mass Squared

If this is a double as the error message suggests, can you maybe treat it at this line as a double and not as a float?

3

Changing the variable type within the code creating the MVAs does not appear to solve the problem unless I have done so wrong. I take it that changing to ‘D’ should be the correct input for setting as a double?

4

Did you change ALL the variables to ‘D’ or just one?

5

The error persists for both where only my “weMissM2__boY4ScleanMask__cm__sp0__bc” variable is set as a double or even with all the variables set as doubles (checking the data types of the variables, they are all doubles).

6

Can you explain how to run your files? I tried but I get: Failed to open TMVA XML file TMVAClassification_Belle_e_id.weights.xml

7

The code in question which raises the error is the Belle_e_tmva_apply.C and I’ve attached a version of the xml weight file to the google drive (the weight file was made with the variable set with float typing).

8

I can reproduce the bug.

Maybe the problem is TMVA::Experimental::Compute<7, float>(e_id_model) which should be changed to double, but then there is a TMVA function that does not support this.

So I guess the workaround is to copy your TTree to one with all variables as float?

Maybe @vpadulan or @moneta can comment.