Learning Symmetry-Independent Jet Representations via Jet-Based Joint Embedding Predictive Architecture

TL;DR

This paper presents J-JEPA, a self-supervised learning method for jet representations in high energy physics that avoids hand-crafted augmentations, enabling versatile applications across multiple tasks.

Contribution

Introduces J-JEPA, a novel jet-based joint embedding predictive architecture that learns invariant representations without hand-crafted augmentations, facilitating cross-task applications.

Findings

J-JEPA achieves competitive performance in jet tagging tasks.

The method avoids biases introduced by manual data augmentations.

J-JEPA provides versatile representations applicable to multiple downstream tasks.

Abstract

In high energy physics, self-supervised learning (SSL) methods have the potential to aid in the creation of machine learning models without the need for labeled datasets for a variety of tasks, including those related to jets -- narrow sprays of particles produced by quarks and gluons in high energy particle collisions. This study introduces an approach to learning jet representations without hand-crafted augmentations using a jet-based joint embedding predictive architecture (J-JEPA), which aims to predict various physical targets from an informative context. As our method does not require hand-crafted augmentation like other common SSL techniques, J-JEPA avoids introducing biases that could harm downstream tasks. Since different tasks generally require invariance under different augmentations, this training without hand-crafted augmentation enables versatile applications, offering a pathway toward a cross-task foundation model. We finetune the representations learned by J-JEPA for jet tagging and benchmark them against task-specific representations.