Overview

This tutorial introduces how to use Umami for more general applications than just jet flavour tagging in the ATLAS experiment.

The basis of this tutorial is the JetClass dataset. It is described in the paper Particle Transformer for Jet Tagging and available here: https://doi.org/10.5281/zenodo.6619768.

Before the dataset can be used for Umami, it needs to be converted in the h5 format with the data structure which is assumed for Umami.

• The preprocessing tutorial section explains how to retrieve the dataset, convert it, and run the umami preprocessing.
• The training tutorial section explains how to run and monitor the training of networks for classification.
• The validation tutorial section explains how to validate the trained model and investigate loss and accuracy as function of training epoch.