Hadronic signals of new-physics origin at the Large Hadron Collider can remain hidden within the copiously produced hadronic jets. Unveiling such signatures require highly performant deep-learning algorithms. We construct a class of Graph Neural Networks (GNN) in the message-passing formalism that makes the network output infra-red and collinear (IRC) safe, an important criterion satisfied within perturbative QCD calculations. Including IRC safety of the network output as a requirement in the construction of the GNN improves its explainability and robustness against theoretical uncertainties in the data. We generalise Energy Flow Networks (EFN), an IRC safe deep-learning algorithm on a point cloud, defining energy weighted local and global readouts on GNNs. Applying the simplest of such networks to identify top quarks, W bosons and quark/gluon jets, we find that it outperforms state-of-the-art EFNs. Additionally, we obtain a general class of graph construction algorithms that give structurally invariant graphs in the IRC limit, a necessary criterion for the IRC safety of the GNN output.
{ "_oai": { "updated": "2022-05-27T00:34:16Z", "id": "oai:repo.scoap3.org:70345", "sets": [ "JHEP" ] }, "authors": [ { "affiliations": [ { "country": "India", "value": "Theoretical Physics Division, Physical Research Laboratory, Shree Pannalal Patel Marg, Ahmedabad, Gujarat, 380009, India", "organization": "Theoretical Physics Division, Physical Research Laboratory" } ], "surname": "Konar", "email": "konar@prl.res.in", "full_name": "Konar, Partha", "given_names": "Partha" }, { "affiliations": [ { "country": "India", "value": "Theoretical Physics Division, Physical Research Laboratory, Shree Pannalal Patel Marg, Ahmedabad, Gujarat, 380009, India", "organization": "Theoretical Physics Division, Physical Research Laboratory" }, { "country": "India", "value": "Discipline of Physics, Indian Institute of Technology, Palaj, Gandhinagar, Gujarat, 382424, India", "organization": "Indian Institute of Technology" } ], "surname": "Ngairangbam", "email": "vishalng@prl.res.in", "full_name": "Ngairangbam, Vishal", "given_names": "Vishal" }, { "affiliations": [ { "country": "UK", "value": "Institute for Particle Physics Phenomenology, Durham University, Durham, DH1 3LE, U.K.", "organization": "Durham University" }, { "country": "UK", "value": "Department of Physics, Durham University, Durham, DH1 3LE, U.K.", "organization": "Durham University" } ], "surname": "Spannowsky", "email": "michael.spannowsky@durham.ac.uk", "full_name": "Spannowsky, Michael", "given_names": "Michael" } ], "titles": [ { "source": "Springer", "title": "Energy-weighted message passing: an infra-red and collinear safe graph neural network algorithm" } ], "dois": [ { "value": "10.1007/JHEP02(2022)060" } ], "publication_info": [ { "page_end": "30", "journal_title": "Journal of High Energy Physics", "material": "article", "journal_volume": "2022", "artid": "JHEP02(2022)060", "year": 2022, "page_start": "1", "journal_issue": "2" } ], "$schema": "http://repo.scoap3.org/schemas/hep.json", "acquisition_source": { "date": "2022-05-27T00:31:36.186425", "source": "Springer", "method": "Springer", "submission_number": "24804aa0dd5411eca1a1661ec451a4e4" }, "page_nr": [ 30 ], "license": [ { "url": "https://creativecommons.org/licenses//by/4.0", "license": "CC-BY-4.0" } ], "copyright": [ { "holder": "The Author(s)", "year": "2022" } ], "control_number": "70345", "record_creation_date": "2022-05-27T00:31:36.186457", "_files": [ { "checksum": "md5:e11d14858c6ec488bb0e9b49ecfffc3e", "filetype": "xml", "bucket": "bcbeb70b-0130-4cce-aa46-f0767174641e", "version_id": "3bbbf23a-a17d-410b-890d-cc9b8dd6bd63", "key": "10.1007/JHEP02(2022)060.xml", "size": 13055 }, { "checksum": "md5:6a4fe6734624e959dd1dbad749ef87e9", "filetype": "pdf/a", "bucket": "bcbeb70b-0130-4cce-aa46-f0767174641e", "version_id": "d7a2b173-18a6-43b7-be61-775cf090cadb", "key": "10.1007/JHEP02(2022)060_a.pdf", "size": 3429377 } ], "collections": [ { "primary": "Journal of High Energy Physics" } ], "arxiv_eprints": [ { "categories": [ "hep-ph" ], "value": "2109.14636" } ], "abstracts": [ { "source": "Springer", "value": "Hadronic signals of new-physics origin at the Large Hadron Collider can remain hidden within the copiously produced hadronic jets. Unveiling such signatures require highly performant deep-learning algorithms. We construct a class of Graph Neural Networks (GNN) in the message-passing formalism that makes the network output infra-red and collinear (IRC) safe, an important criterion satisfied within perturbative QCD calculations. Including IRC safety of the network output as a requirement in the construction of the GNN improves its explainability and robustness against theoretical uncertainties in the data. We generalise Energy Flow Networks (EFN), an IRC safe deep-learning algorithm on a point cloud, defining energy weighted local and global readouts on GNNs. Applying the simplest of such networks to identify top quarks, W bosons and quark/gluon jets, we find that it outperforms state-of-the-art EFNs. Additionally, we obtain a general class of graph construction algorithms that give structurally invariant graphs in the IRC limit, a necessary criterion for the IRC safety of the GNN output." } ], "imprints": [ { "date": "2022-02-08", "publisher": "Springer" } ] }