PELICAN-lite documentation

This is an efficient reimplementation of the PELICAN architecture. PELICAN was first published at the ML4PS workshop 2022 and on JHEP. The official implementation is available on https://github.com/abogatskiy/PELICAN.

This implementation aims to improve efficiency and ease of use. For toptagging with batch size 100, we find 8x reduced memory usage and a 3x training speedup compared to the original implementation. PELICAN-lite can be used as the Frames-Net in Lorentz Local Canonicalization (LLoCa).

• Quickstart

• Differences compared to the original PELICAN implementation

• API Reference

Citation

If you find this package useful, please cite these papers:

@article{Favaro:2025pgz,
   author = "Favaro, Luigi and Gerhartz, Gerrit and Hamprecht, Fred A. and Lippmann, Peter and Pitz, Sebastian and Plehn, Tilman and Qu, Huilin and Spinner, Jonas",
   title = "{Lorentz-Equivariance without Limitations}",
   eprint = "2508.14898",
   archivePrefix = "arXiv",
   primaryClass = "hep-ph",
   month = "8",
   year = "2025"
}
@article{Bogatskiy:2023nnw,
   author = "Bogatskiy, Alexander and Hoffman, Timothy and Miller, David W. and Offermann, Jan T. and Liu, Xiaoyang",
   title = "{Explainable equivariant neural networks for particle physics: PELICAN}",
   eprint = "2307.16506",
   archivePrefix = "arXiv",
   primaryClass = "hep-ph",
   doi = "10.1007/JHEP03(2024)113",
   journal = "JHEP",
   volume = "03",
   pages = "113",
   year = "2024"
}
@article{Bogatskiy:2022czk,
   author = "Bogatskiy, Alexander and Hoffman, Timothy and Miller, David W. and Offermann, Jan T.",
   title = "{PELICAN: Permutation Equivariant and Lorentz Invariant or Covariant Aggregator Network for Particle Physics}",
   eprint = "2211.00454",
   archivePrefix = "arXiv",
   primaryClass = "hep-ph",
   month = "11",
   year = "2022"
}