Using Neural Networks to analyze Higgs Boson Data

Github Link

The goal of the Higgs Boson Machine Learning Challenge is to explore the potential of advanced machine learning methods to improve the discovery significance of the experiment.  Using simulated data with features characterizing events detected by ATLAS, we classified events into “tau tau decay of a Higgs boson” versus “background.” 

The first 10.5 MM rows are trained and validated with a 9:1 ratio and the last 500 rows are tested for accuracy, precision, recall and ROC score metrics, not unlike the classification models you will see on this site. However, using tools such as Tensorflow, we can optimize parameters for number of layers, neurons (units), activation function, batch size, kernel initializer, optimizer, learning rate, epsilon, decay rate, and dropout rate to get the greatest classification accuracy.