Exploring Machine Learning for Particle Physics

Abstract

In this article, we report our work on the Kaggle Challenge: Flavours of Physics (2016). The main goal of this challenge is to develop powerful classifiers for the detection of “new physics” – specifically, violation the lepton flavour conservation guaranteed by the standard model – in the Large Hadron Collider (LHC). We build several models for the challenge, which include boosted decision trees and neural networks, and report our findings on their performance. We include a cautionary tale of how the use of some specific features of the data can dramatically impact the physical relevance of the results.

Type
Publication
Technical report, 2017
Aristide Baratin
Aristide Baratin
PhD Candidate

Related