Deep Learning From Four Vectors

TL;DR

This paper demonstrates how deep neural networks operating on four-vectors can outperform traditional methods in particle physics data analysis, highlighting architecture design and physics-informed approaches.

Contribution

It introduces deep learning techniques for four-vector data, including architecture extensions and methods to incorporate physics knowledge, advancing particle physics data analysis.

Findings

Deep networks outperform shallow models on four-vector data

Extensions to parameterized networks improve flexibility

Incorporating physics knowledge enhances model performance

Abstract

An early example of the ability of deep networks to improve the statistical power of data collected in particle physics experiments was the demonstration that such networks operating on lists of particle momenta (four-vectors) could outperform shallow networks using features engineered with domain knowledge. A benchmark case is described, with extensions to parameterized networks. A discussion of data handling and architecture is presented, as well as a description of how to incorporate physics knowledge into the network architecture.