2000 character limit reached
19 Parameters Is All You Need: Tiny Neural Networks for Particle Physics (2310.16121v3)
Published 24 Oct 2023 in hep-ph, cs.LG, and hep-ex
Abstract: As particle accelerators increase their collision rates, and deep learning solutions prove their viability, there is a growing need for lightweight and fast neural network architectures for low-latency tasks such as triggering. We examine the potential of one recent Lorentz- and permutation-symmetric architecture, PELICAN, and present its instances with as few as 19 trainable parameters that outperform generic architectures with tens of thousands of parameters when compared on the binary classification task of top quark jet tagging.
- Georges Aad “Performance of the ATLAS Level-1 topological trigger in Run 2” In Eur. Phys. J. C 82.1, 2022, pp. 7 DOI: 10.1140/epjc/s10052-021-09807-0
- “Lorentz Group Equivariant Neural Network for Particle Physics” In ICML 2020 ICML, 2020 DOI: 10.48550/arXiv.2006.04780
- “PELICAN: Permutation Equivariant and Lorentz Invariant or Covariant Aggregator Network for Particle Physics”, 2022 DOI: 10.48550/arXiv.2211.00454
- “Explainable Equivariant Neural Networks for Particle Physics: PELICAN”, 2023 arXiv:2307.16506 [hep-ph]
- W Buttinger “The ATLAS Level-1 Trigger System” On behalf of the ATLAS collaboration, 2012 DOI: 10.1088/1742-6596/396/1/012010
- Matteo Cacciari, Gavin P. Salam and Gregory Soyez “The anti-ktsubscript𝑘𝑡k_{t}italic_k start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT jet clustering algorithm” In JHEP 04, 2008, pp. 063 DOI: 10.1088/1126-6708/2008/04/063
- Ranit Das, Gregor Kasieczka and David Shih “Feature Selection with Distance Correlation”, 2022 arXiv:2212.00046 [hep-ph]
- Stephen D. Ellis, Zoltan Kunszt and Davison E. Soper “Jets at hadron colliders at order α−s3:𝛼superscript𝑠:3absent\alpha-s^{3:}italic_α - italic_s start_POSTSUPERSCRIPT 3 : end_POSTSUPERSCRIPT A Look inside” In Phys. Rev. Lett. 69, 1992, pp. 3615–3618 DOI: 10.1103/PhysRevLett.69.3615
- “An efficient Lorentz equivariant graph neural network for jet tagging” In JHEP 2022.7, 2022, pp. 30 DOI: 10.1007/JHEP07(2022)030
- “Top Quark Tagging Reference Dataset”, 2019 URL: https://zenodo.org/record/2603256
- “Converted Top Tagging Dataset”, 2020 URL: https://osf.io/7u3fk/?view_only=8c42f1b112ab4a43bcf208012f9db2df
- “The Machine Learning landscape of top taggers” In SciPost Phys. 7 SciPost, 2019, pp. 014 DOI: 10.21468/SciPostPhys.7.1.014
- Patrick T. Komiske, Eric M. Metodiev and Jesse Thaler “Energy flow polynomials: A complete linear basis for jet substructure” In JHEP 04, 2018, pp. 013 DOI: 10.1007/JHEP04(2018)013
- Patrick T. Komiske, Eric M. Metodiev and Jesse Thaler “Energy flow networks: deep sets for particle jets” In JHEP 2019.1, 2019, pp. 121 DOI: 10.1007/jhep01(2019)121
- “Does Lorentz-symmetric design boost network performance in jet physics?”, 2022 arXiv:2208.07814 [hep-ph]
- “Invariant and Equivariant Graph Networks”, 2018 arXiv: http://arxiv.org/abs/1812.09902
- “JEDI-net: a jet identification algorithm based on interaction networks” In Eur. Phys. J. C 80.1, 2020, pp. 58 DOI: 10.1140/epjc/s10052-020-7608-4
- Jose M Munoz, Ilyes Batatia and Christoph Ortner “Boost invariant polynomials for efficient jet tagging” In Machine Learning: Science and Technology 3.4 IOP Publishing, 2022, pp. 04LT05 DOI: 10.1088/2632-2153/aca9ca
- “Permutation Equivariant Layers for Higher Order Interactions” In AISTATS PMLR, 2022, pp. 5987–6001 URL: https://proceedings.mlr.press/v151/pan22a.html
- “Jet Constituents for Deep Neural Network Based Top Quark Tagging”, 2017 arXiv:1704.02124 [hep-ex]
- “Jet tagging via particle clouds” In Phys. Rev. D 101.5, 2020 DOI: 10.1103/PhysRevD.101.056019
- Huilin Qu, Congqiao Li and Sitian Qian “Particle Transformer for Jet Tagging”, 2022 arXiv:2202.03772 [hep-ph]
- “Optimizing Graph Neural Networks for Jet Tagging in Particle Physics on FPGAs” In 2022 32nd International Conference on Field-Programmable Logic and Applications (FPL), 2022, pp. 327–333 DOI: 10.1109/FPL57034.2022.00057
- “LL-GNN: Low Latency Graph Neural Networks on FPGAs for High Energy Physics”, 2023 arXiv:2209.14065 [cs.AR]
- Jesse Thaler and Ken Van Tilburg “Identifying Boosted Objects with N-subjettiness” In JHEP 03, 2011, pp. 015 DOI: 10.1007/JHEP03(2011)015