README.md

Higgs ML (Higgs boson Machine Learning Classifiers)

Supervised Machine Learning techniques used to categorise VH, H(bb) Higgs boson decay events using data collected from the Large Hadron Collider, CERN.

Repo Contents

1. Boosted Decision Trees

a) Different BDT Classifiers

• XGBoostBDT.py – XGBoost BDT Classifier
• AdaBoostBDT.py – AdaBoost Boosted Decision Tree Classifier
• randomForestBDT.py – Random Forest Classifier

b) XGBoost Optimisation

• variationOfOptimalSensitivity.py – Determines optimal Input Variables that affect Classifier Sensitivity
• XGBoostOptimisation.py – Determines optimal hyperparameters of XGBoost Classifier
• XGBoostStandardError.py – Calculates standard error of the XGBoost Classifier

c) Dataset Analysis

• datasetCorrelationMatrices.py – Plots correlation matrices of dataset of model's input variables

2. Deep Neural Networks

• DNN.py – DNN Classifier
• DNNOptimisation.py – Determines Optimal Hyperparameters of DNN Classifier
• DNNStandardError.py – Calculates Standard Error of the DNN Classifier

3. Adversarial Neural Networks

a) Dijet-mass Aware ANN

• dijetmassANN.py – Dijet-mass Aware Adversarial Neural Network Classifier
• dijetmassANN_withmBBDistributionGraphs.py – Dijet-mass Aware Adversarial Neural Network Classifier with Graphs

b) Monte Carlo Generator Aware ANN

• generatorANN.py – Monte Carlo Event Generator Aware Adversarial Neural Network Classifier
• generatorANN_withmBBDistributionGraphs.py – Monte Carlo Event Generator Aware Adversarial Neural Network Classifier with Graphs

4. Dataset & Plotting

a) H(bb) Monte Carlo Generated Event Dataset

• CSV – Original Dataset. Used for training Boosted Decision Tree Classifiers and Deep Neural Network Classifiers.
• CSV_withBinnedDijetMassValues – Original Dataset, with added column of data which bins the raw dijet mass values into 10 equally sized bins (labelled 1 to 10). Used for training Dijet mass-aware Adversarial Neural Network Classifiers.
• CSV_differentGenerators – Original Dataset, which has been duplicated and a non-linear distortion applied to one of the copies, thus simulating Data originating from different Monte Carlo Generators. Used for training Generator-aware Adversarial Neural Networks Classifiers.

b) Graph Plotting Files

• bdtPlotting.py – Scripts used to plot Boosted Decision Tree (BDT) Output Graphs.
• nnPlotting.py – Scripts used to plot Neural Network (NN) and Adversarial Neural Network (ANN) Output Graphs.
• sensitivity.py – Scripts used to calculate sensitivity of BDT and NN and ANN Classifiers.

Download Repo

• Download Link is HERE

• Fork Repo using:

$ git clone https://github.com/louisheery/higgs-ML.git