pybdt

This project provides a Python interface for signal/background discrimination using Boosted Decision Tree classifiers (BDTs).

This project is maintained by Mike Richman <mike.d.richman@gmail.com>.

Note that the example scripts are rather complete as well as complementary to this documentation. Feel free to email me with any questions.

• Release Notes

• pybdt User Manual
  • Statement of the problem
  • The ABC example
    • Input variable distributions
  • Introduction to decision tree classifiers
    • Classification example
    • Some definitions
    • Scoring using purity information
    • Decision tree training
  • Introduction to boosted decision tree classifiers
    • AdaBoost
    • Calculating the BDT score
    • Example BDT score distribution
  • Introduction to overtraining
    • Training sample overtraining
    • Data/Monte Carlo overtraining
  • Introduction to pruning
    • Same leaf pruning
    • Cost complexity pruning
  • Introduction to decision tree randomization
  • DataSets and Training
    • DataSet objects
    • Training
  • Validator Setup
    • Validator Construction
    • DataSet weighting
  • Validator Usage
    • Summary Plots
    • Overtraining Check
    • Other Plotting Features
  • Tree and Forest Inspection
    • Direct Acces
    • Visualization
    • Variable Importance
  • Classifier Application
    • Scoring with IceTray
    • Scoring without IceTray
  • Global Epsilon Reference
• Common Pitfalls
  • Outliers
  • Energy Proxies
  • X server unavailable
• pybdt package
  • Submodules
    • pybdt.convertdsmodule module
    • pybdt.histlite module
    • pybdt.ml module
    • pybdt.util module
    • pybdt.validate module
    • pybdt.viz module