Github repository for DarkCLR, a framework for the detection of semi-visible jets.
This is the reference codebase for:
Semi-visible jets, energy-based models, and self-supervision
arXiv:2312.03067
Favaro L., Kraemer M., Modak T., Plehn T., Rueschkamp J.
The QCD background and the main ''Aachen'' signal datasets are published on Zenodo:
Usage:
TransformerEncoder.py includes the transformer class
contrastive_losses.py defines the CLR loss function
jet_augmentations.py contains the physical and anomalous jet augmentations
We provide an example run script with the set of default parameters in run_scripts.
Related refs.
Dillon B. et al.
Anomalies, Representations, and Self-Supervision
arXiv:2301.04660
Dillon B. et al.
Symmetries, Safety, and Self-supervision
arXiv:2108.04253
Dillon B. et al.
A normalized autoencoder for LHC triggers
arXiv:2206.14225