The Toolkit for Multivariate Data Analysis with ROOT (TMVA) is a ROOT-integrated project providing a machine learning environment for the processing and evaluation of multivariate classification, both binary and multi class, and regression techniques targeting applications in high-energy physics. The package includes:

• Neural networks
  • Deep networks
  • Multilayer perceptron
• Boosted/Bagged decision trees
• Function discriminant analysis (FDA)
• Linear discriminant analysis (H-Matrix and Fisher discriminants)
• Multidimensional probability density estimation (PDE - range-search approach)
• Multidimensional k-nearest neighbour classifier
• Predictive learning via rule ensembles (RuleFit)
• Projective likelihood estimation (PDE approach)
• Rectangular cut optimisation
• Support Vector Machine (SVM)

Tutorials

Online examples of how to use TMVA in both Python and C++ can be found at https://swan.web.cern.ch/content/machine-learning. In particular

• C++
  • Classification/Regression. Simple setups for classification and regression.
  • Cross Validation. Shows how cross validation is used in TMVA.
• Python
  • DNN. How to train a deep neural network with the TMVA backend.
  • Keras with tmva. Define a keras model and train through TMVA.
  • Jupyter integration. Shows functionality available in Jupyter notebooks. On your local machine you can start a jupyter notebook with root -l --notebook.

There are also tutorials available with your ROOT distribution at $ROOTSYS/tutorials/tmva where $ROOTSYS is the path to your ROOT installation.

Links

• For an expanded introduction to TMVA, check the executive summary.
• TMVA comes with a very good user guide for in-depth usage and information.
• A quick start to quickly get up and running can be found here
• For further questions, you are welcome to contact us through the ROOT forum.