Differentiable Vertex Fitting for Jet Flavour Tagging

TL;DR

This paper introduces a differentiable vertex fitting algorithm that integrates into neural networks for jet flavour tagging, enabling end-to-end training and improved classification of heavy flavour jets in high energy physics.

Contribution

It presents a novel differentiable vertex fitting method that incorporates physics knowledge into neural networks for enhanced jet flavour tagging.

Findings

Improved heavy flavour jet classification accuracy.

Seamless integration of vertex fitting into transformer-based models.

Demonstrated benefits of differentiable programming in physics applications.

Abstract

We propose a differentiable vertex fitting algorithm that can be used for secondary vertex fitting, and that can be seamlessly integrated into neural networks for jet flavour tagging. Vertex fitting is formulated as an optimization problem where gradients of the optimized solution vertex are defined through implicit differentiation and can be passed to upstream or downstream neural network components for network training. More broadly, this is an application of differentiable programming to integrate physics knowledge into neural network models in high energy physics. We demonstrate how differentiable secondary vertex fitting can be integrated into larger transformer-based models for flavour tagging and improve heavy flavour jet classification.