How to train/test with tree contains TClonesArray

TMVA
1

Dear experts,

I know how to use TMVA/PyMVA for training/testing with simple tree structure for MultiClass Classification. But if I have a tree that contains TClonesArray of variables that I want to train on (event loop and track loop). How to extract the variables from the tree ?

#include <iostream>

#include "TString.h"
#include "TFile.h"
#include "TTree.h"
#include "TSystem.h"
#include "TROOT.h"
#include "TMVA/Factory.h"
#include "TMVA/Reader.h"
#include "TMVA/DataLoader.h"
#include "TMVA/PyMethodBase.h"

TString pythonSrc = "\
from keras.models import Sequential\n\
from keras.layers.core import Dense, Dropout, Activation\n\
from keras.optimizers import Adam\n\
\n\
model = Sequential()\n\
model.add(Dense(100, activation=\"tanh\", input_dim=26))\n\
model.add(Dense(50, activation=\"tanh\"))\n\
model.add(Dense(3, activation=\"softmax\"))\n\
model.compile(loss=\"categorical_crossentropy\", optimizer=Adam(), metrics=[\"accuracy\",])\n\
model.summary()\n\
model.save(\"kerasModelMulticlass.h5\")\n";

int PyKerasMulticlass(){
   // Get data file
   TFile *input0 = TFile::Open("./TrainingSet/electrons_tree.root");
   TFile *input1 = TFile::Open("./TrainingSet/pions_tree.root");
   TFile *input2 = TFile::Open("./TrainingSet/muons_tree.root");

   // Build model from python file
   std::cout << "Generate keras model..." << std::endl;
   UInt_t ret;
   ret = gSystem->Exec("echo '"+pythonSrc+"' > generateKerasModelMulticlass.py");
   if(ret!=0){
       std::cout << "[ERROR] Failed to write python code to file" << std::endl;
       return 1;
   }
   ret = gSystem->Exec("ipython generateKerasModelMulticlass.py");
   if(ret!=0){
       std::cout << "[ERROR] Failed to generate model using python" << std::endl;
       return 1;
   }

   // Setup PyMVA and factory
   std::cout << "Setup TMVA..." << std::endl;
   TMVA::PyMethodBase::PyInitialize();
   TFile* outputFile = TFile::Open("ResultsTestPyKerasMulticlass.root", "RECREATE");
   TMVA::Factory *factory = new TMVA::Factory("testPyKerasMulticlass", outputFile,
      "!V:Silent:Color:!DrawProgressBar:Transformations=None:AnalysisType=multiclass");

   // Load data
   TMVA::DataLoader *dataloader = new TMVA::DataLoader("datasetTestPyKerasMulticlass");

   TTree *signal0 = (TTree*)input0->Get("t1");
   TTree *signal1 = (TTree*)input1->Get("t1");
   TTree *signal2 = (TTree*)input2->Get("t1");

   dataloader->AddTree(signal0, "Electrons");
   dataloader->AddTree(signal1, "Pions");
   dataloader->AddTree(signal2, "Muons");

    TClonesArray *track = new TClonesArray("hTrack");
    signal0->GetBranch("track")->SetAutoDelete(kFALSE);
    signal0->SetBranchAddress("hTrack",&track);

   // Is this the right way to do it ???!!
   dataloader->AddVariable("track.fXmomentum", &track->fXmomentum);


   dataloader->PrepareTrainingAndTestTree("",
      "SplitMode=Random:NormMode=NumEvents:!V");

   // Book and train method
   factory->BookMethod(dataloader, TMVA::Types::kPyKeras, "PyKeras",
      "!H:!V:VarTransform=D,G:FilenameModel=kerasModelMulticlass.h5:FilenameTrainedModel=trainedKerasModelMulticlass.h5:NumEpochs=10:BatchSize=128:SaveBestOnly=false:Verbose=1");
   std::cout << "Train model..." << std::endl;
   factory->TrainAllMethods();
   std::cout << "Test model..." << std::endl;
   //factory->TestAllMethods();
   std::cout << "Evaluate model..." << std::endl;
   //factory->EvaluateAllMethods();   

   // Clean-up
   delete factory;
   delete dataloader;
   delete outputFile;
   return 0;
}

int main(){
   int err = PyKerasMulticlass();
   return err;
}

Thank you in advance

2

Hi,
Currently in TMVA we support only flat data structures as input branch of the TTree and not collections as TClonesArray. Using for example the RDataFrame you should be able to create a cloned trees with the structure you want and you can use it as input to TMVA

Lorenzo

1 Like