Papers
Topics
Authors
Recent
2000 character limit reached

Comparing Point Cloud Strategies for Collider Event Classification (2212.10659v2)

Published 20 Dec 2022 in hep-ph, hep-ex, and physics.data-an

Abstract: In this paper, we compare several event classification architectures defined on the point cloud representation of collider events. These approaches, which are based on the frameworks of deep sets and edge convolutions, circumvent many of the difficulties associated with traditional feature engineering. To benchmark our architectures against more traditional event classification strategies, we perform a case study involving Higgs boson decays to tau leptons. We find a 2.5 times increase in performance compared to a baseline ATLAS analysis with engineered features. Our point cloud architectures can be viewed as simplified versions of graph neural networks, where each particle in the event corresponds to a graph node. In our case study, we find the best balance of performance and computational cost for simple pairwise architectures, which are based on learned edge features.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (67)
  1. CMS Collaboration, “Measurement of t⁢t¯⁢Ht¯tH\mathrm{t\overline{t}H}roman_t over¯ start_ARG roman_t end_ARG roman_H production in the H→b⁢b¯→Hb¯b\mathrm{H\rightarrow b\overline{b}}roman_H → roman_b over¯ start_ARG roman_b end_ARG decay channel in 41.5⁢fb−141.5superscriptfb141.5\,\mathrm{fb}^{-1}41.5 roman_fb start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT of proton-proton collision data at s=13⁢TeV𝑠13TeV\sqrt{s}=13\,\mathrm{TeV}square-root start_ARG italic_s end_ARG = 13 roman_TeV,”   (2019).
  2. Albert M Sirunyan et al. (CMS), “Search for t⁢t¯⁢Ht¯tH\mathrm{t}\overline{\mathrm{t}}\mathrm{H}roman_t over¯ start_ARG roman_t end_ARG roman_H production in the H→b⁢b¯→Hb¯b\mathrm{H}\to\mathrm{b}\overline{\mathrm{b}}roman_H → roman_b over¯ start_ARG roman_b end_ARG decay channel with leptonic t⁢t¯t¯t\mathrm{t}\overline{\mathrm{t}}roman_t over¯ start_ARG roman_t end_ARG decays in proton-proton collisions at s=13𝑠13\sqrt{s}=13square-root start_ARG italic_s end_ARG = 13 TeV,” JHEP 03, 026 (2019), arXiv:1804.03682 [hep-ex] .
  3. Georges Aad et al. (ATLAS), “Measurements of Higgs boson production cross-sections in the H→τ+⁢τ−→𝐻superscript𝜏superscript𝜏H\to\tau^{+}\tau^{-}italic_H → italic_τ start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT italic_τ start_POSTSUPERSCRIPT - end_POSTSUPERSCRIPT decay channel in pp collisions at s𝑠\sqrt{s}square-root start_ARG italic_s end_ARG = 13 TeV with the ATLAS detector,” JHEP 08, 175 (2022), arXiv:2201.08269 [hep-ex] .
  4. Georges Aad et al. (ATLAS), “Test of CP invariance in vector-boson fusion production of the Higgs boson in the H → τ𝜏\tauitalic_ττ𝜏\tauitalic_τ channel in proton–proton collisions at s=13TeV with the ATLAS detector,” Phys. Lett. B 805, 135426 (2020), arXiv:2002.05315 [hep-ex] .
  5. Morad Aaboud et al. (ATLAS), “Cross-section measurements of the Higgs boson decaying into a pair of τ𝜏\tauitalic_τ-leptons in proton-proton collisions at s=13𝑠13\sqrt{s}=13square-root start_ARG italic_s end_ARG = 13 TeV with the ATLAS detector,” Phys. Rev. D 99, 072001 (2019), arXiv:1811.08856 [hep-ex] .
  6. M. Aaboud et al. (ATLAS), “Evidence for the H→b⁢b¯→𝐻𝑏¯𝑏H\to b\overline{b}italic_H → italic_b over¯ start_ARG italic_b end_ARG decay with the ATLAS detector,” JHEP 12, 024 (2017), arXiv:1708.03299 [hep-ex] .
  7. Morad Aaboud et al. (ATLAS), “Observation of H→b⁢b¯→𝐻𝑏¯𝑏H\rightarrow b\bar{b}italic_H → italic_b over¯ start_ARG italic_b end_ARG decays and V⁢H𝑉𝐻VHitalic_V italic_H production with the ATLAS detector,” Phys. Lett. B 786, 59–86 (2018a), arXiv:1808.08238 [hep-ex] .
  8. Morad Aaboud et al. (ATLAS), “Search for the standard model Higgs boson produced in association with top quarks and decaying into a b⁢b¯𝑏¯𝑏b\bar{b}italic_b over¯ start_ARG italic_b end_ARG pair in p⁢p𝑝𝑝ppitalic_p italic_p collisions at s𝑠\sqrt{s}square-root start_ARG italic_s end_ARG = 13 TeV with the ATLAS detector,” Phys. Rev. D 97, 072016 (2018b), arXiv:1712.08895 [hep-ex] .
  9. Georges Aad et al. (ATLAS), “Measurements of W⁢H𝑊𝐻WHitalic_W italic_H and Z⁢H𝑍𝐻ZHitalic_Z italic_H production in the H→b⁢b¯→𝐻𝑏¯𝑏H\rightarrow b\bar{b}italic_H → italic_b over¯ start_ARG italic_b end_ARG decay channel in p⁢p𝑝𝑝ppitalic_p italic_p collisions at 13 TeV with the ATLAS detector,” Eur. Phys. J. C 81, 178 (2021), arXiv:2007.02873 [hep-ex] .
  10. Georges Aad et al. (ATLAS), “Search for the b⁢b¯𝑏¯𝑏b\bar{b}italic_b over¯ start_ARG italic_b end_ARG decay of the Standard Model Higgs boson in associated (W/Z)⁢H𝑊𝑍𝐻(W/Z)H( italic_W / italic_Z ) italic_H production with the ATLAS detector,” JHEP 01, 069 (2015), arXiv:1409.6212 [hep-ex] .
  11. Manzil Zaheer, Satwik Kottur, Siamak Ravanbhakhsh, Barnabás Póczos, Ruslan Salakhutdinov,  and Alexander J. Smola, “Deep sets,” Advances in Neural Information Processing Systems 2017-Decem, 3392–3402 (2017), arXiv:1703.06114 .
  12. Patrick T. Komiske, Eric M. Metodiev,  and Jesse Thaler, “Energy Flow Networks: Deep Sets for Particle Jets,” JHEP 01, 121 (2019), arXiv:1810.05165 [hep-ph] .
  13. Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E. Sarma, Michael M. Bronstein,  and Justin M. Solomon, “Dynamic Graph CNN for Learning on Point Clouds,”   (2018), arXiv:1801.07829 [cs.CV] .
  14. Huilin Qu and Loukas Gouskos, “ParticleNet: Jet Tagging via Particle Clouds,” Phys. Rev. D 101, 056019 (2020), arXiv:1902.08570 [hep-ph] .
  15. Matthew J. Dolan and Ayodele Ore, “Equivariant Energy Flow Networks for Jet Tagging,” Phys. Rev. D 103, 074022 (2021), arXiv:2012.00964 [hep-ph] .
  16. Serguei Chatrchyan et al. (CMS), ‘‘Evidence for the 125 GeV Higgs boson decaying to a pair of τ𝜏\tauitalic_τ leptons,” JHEP 05, 104 (2014), arXiv:1401.5041 [hep-ex] .
  17. Albert M Sirunyan et al. (CMS), “Observation of the Higgs boson decay to a pair of τ𝜏\tauitalic_τ leptons with the CMS detector,” Phys. Lett. B 779, 283–316 (2018), arXiv:1708.00373 [hep-ex] .
  18. Armen Tumasyan et al. (CMS), “Measurement of the inclusive and differential Higgs boson production cross sections in the decay mode to a pair of τ𝜏\tauitalic_τ leptons in pp collisions at s=𝑠absent\sqrt{s}=square-root start_ARG italic_s end_ARG = 13 TeV,” Phys. Rev. Lett. 128, 081805 (2022), arXiv:2107.11486 [hep-ex] .
  19. “Measurements of Higgs boson production in the decay channel with a pair of τ𝜏\tauitalic_τ leptons in proton-proton collisions at s𝑠\sqrt{s}square-root start_ARG italic_s end_ARG = 13 TeV,”  (2022), arXiv:2204.12957 [hep-ex] .
  20. Eric A. Moreno, Olmo Cerri, Javier M. Duarte, Harvey B. Newman, Thong Q. Nguyen, Avikar Periwal, Maurizio Pierini, Aidana Serikova, Maria Spiropulu,  and Jean-Roch Vlimant, “JEDI-net: a jet identification algorithm based on interaction networks,” Eur. Phys. J. C 80, 58 (2020), arXiv:1908.05318 [hep-ex] .
  21. Vinicius Mikuni and Florencia Canelli, “ABCNet: An attention-based method for particle tagging,” Eur. Phys. J. Plus 135, 463 (2020), arXiv:2001.05311 [physics.data-an] .
  22. Amit Chakraborty, Sung Hak Lim, Mihoko M. Nojiri,  and Michihisa Takeuchi, “Neural Network-based Top Tagger with Two-Point Energy Correlations and Geometry of Soft Emissions,” JHEP 07, 111 (2020), arXiv:2003.11787 [hep-ph] .
  23. Jonathan Shlomi, Peter Battaglia,  and Jean-Roch Vlimant, “Graph Neural Networks in Particle Physics,”   (2020), 10.1088/2632-2153/abbf9a, arXiv:2007.13681 [hep-ex] .
  24. Javier Duarte and Jean-Roch Vlimant, “Graph Neural Networks for Particle Tracking and Reconstruction,”   (2020), 10.1142/9789811234033_0012, arXiv:2012.01249 [hep-ph] .
  25. Savannah Thais, Paolo Calafiura, Grigorios Chachamis, Gage DeZoort, Javier Duarte, Sanmay Ganguly, Michael Kagan, Daniel Murnane, Mark S. Neubauer,  and Kazuhiro Terao, “Graph Neural Networks in Particle Physics: Implementations, Innovations, and Challenges,” in 2022 Snowmass Summer Study (2022) arXiv:2203.12852 [hep-ex] .
  26. Delon Shen, “Classifying Collider Events with Point Clouds,” https://github.com/DelonShen/classifying-collider-events-with-point-clouds (2022).
  27. Anja Butter et al., “The Machine Learning landscape of top taggers,” SciPost Phys. 7, 014 (2019), arXiv:1902.09914 [hep-ph] .
  28. A. Djouadi, J. Kalinowski,  and M. Spira, “HDECAY: A Program for Higgs boson decays in the standard model and its supersymmetric extension,” Comput. Phys. Commun. 108, 56–74 (1998), arXiv:hep-ph/9704448 .
  29. D. de Florian et al. (LHC Higgs Cross Section Working Group), “Handbook of LHC Higgs Cross Sections: 4. Deciphering the Nature of the Higgs Sector,”  2/2017 (2016), 10.23731/CYRM-2017-002, arXiv:1610.07922 [hep-ph] .
  30. Albert M Sirunyan et al. (CMS), “Measurement of the Higgs boson production rate in association with top quarks in final states with electrons, muons, and hadronically decaying tau leptons at s=𝑠absent\sqrt{s}=square-root start_ARG italic_s end_ARG = 13 TeV,” Eur. Phys. J. C 81, 378 (2021), arXiv:2011.03652 [hep-ex] .
  31. R. Keith Ellis, I. Hinchliffe, M. Soldate,  and J. J. van der Bij, “Higgs Decay to tau+ tau-: A Possible Signature of Intermediate Mass Higgs Bosons at the SSC,” Nucl. Phys. B 297, 221–243 (1988).
  32. A. Elagin, P. Murat, A. Pranko,  and A. Safonov, “A New Mass Reconstruction Technique for Resonances Decaying to di-tau,” Nucl. Instrum. Meth. A 654, 481–489 (2011), arXiv:1012.4686 [hep-ex] .
  33. Partha Konar and Abhaya Kumar Swain, “Reconstructing semi-invisible events in resonant tau pair production from Higgs,” Phys. Lett. B 757, 211–215 (2016), arXiv:1602.00552 [hep-ph] .
  34. Johan Alwall, Michel Herquet, Fabio Maltoni, Olivier Mattelaer,  and Tim Stelzer, “MadGraph 5 : Going Beyond,” JHEP 06, 128 (2011), arXiv:1106.0522 [hep-ph] .
  35. Torbjörn Sjöstrand, Stefan Ask, Jesper R. Christiansen, Richard Corke, Nishita Desai, Philip Ilten, Stephen Mrenna, Stefan Prestel, Christine O. Rasmussen,  and Peter Z. Skands, “An introduction to PYTHIA 8.2,” Comput. Phys. Commun. 191, 159–177 (2015), arXiv:1410.3012 [hep-ph] .
  36. J. de Favereau, C. Delaere, P. Demin, A. Giammanco, V. Lemaître, A. Mertens,  and M. Selvaggi (DELPHES 3), “DELPHES 3, A modular framework for fast simulation of a generic collider experiment,” JHEP 02, 057 (2014), arXiv:1307.6346 [hep-ex] .
  37. Matteo Cacciari, Gavin P. Salam,  and Gregory Soyez, “The anti-ktsubscript𝑘𝑡k_{t}italic_k start_POSTSUBSCRIPT italic_t end_POSTSUBSCRIPT jet clustering algorithm,” JHEP 04, 063 (2008), arXiv:0802.1189 [hep-ph] .
  38. Matteo Cacciari, Gavin P. Salam,  and Gregory Soyez, “FastJet User Manual,” Eur. Phys. J. C 72, 1896 (2012), arXiv:1111.6097 [hep-ph] .
  39. Jason Gallicchio and Matthew D. Schwartz, “Quark and Gluon Jet Substructure,” JHEP 04, 090 (2013), arXiv:1211.7038 [hep-ph] .
  40. Murray Rosenblatt, “Remarks on Some Nonparametric Estimates of a Density Function,” The Annals of Mathematical Statistics 27, 832–837 (1956).
  41. Emanuel Parzen, “On Estimation of a Probability Density Function and Mode,” The Annals of Mathematical Statistics 33, 1065–1076 (1962).
  42. Laurens Van Der Maaten, “Learning a parametric embedding by preserving local structure,” Journal of Machine Learning Research 5, 384–391 (2009).
  43. Laurens Van Der Maaten and Geoffrey Hinton, “Visualizing non-metric similarities in multiple maps,” Machine Learning 87, 33–55 (2012).
  44. Carlos R. García-Alonso, Leonor M. Pérez-Naranjo,  and Juan C. Fernández-Caballero, “Multiobjective evolutionary algorithms to identify highly autocorrelated areas: The case of spatial distribution in financially compromised farms,” Annals of Operations Research 219, 187–202 (2014).
  45. Laurens Van Der Maaten, “Accelerating t-SNE using tree-based algorithms,” Journal of Machine Learning Research 15, 3221–3245 (2015).
  46. S. Peleg, M. Werman,  and H. Rom, “A unified approach to the change of resolution: space and gray-level,” IEEE Transactions on Pattern Analysis and Machine Intelligence 11, 739–742 (1989).
  47. Y. Rubner, C. Tomasi,  and L.J. Guibas, “A metric for distributions with applications to image databases,” in Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271) (Narosa Publishing House) pp. 59–66.
  48. Yossi Rubner, Carlo Tomasi,  and Leonidas J. Guibas, “The earth mover’s distance as a metric for image retrieval,” Int. J. Comput. Vision 40, 99–121 (2000).
  49. Anja Butter, Gregor Kasieczka, Tilman Plehn,  and Michael Russell, “Deep-learned Top Tagging with a Lorentz Layer,” SciPost Phys. 5, 028 (2018), arXiv:1707.08966 [hep-ph] .
  50. M. Erdmann, E. Geiser, Y. Rath,  and M. Rieger, “Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering,” JINST 14, P06006 (2019), arXiv:1812.09722 [hep-ex] .
  51. Alexander Bogatskiy, Brandon Anderson, Jan T. Offermann, Marwah Roussi, David W. Miller,  and Risi Kondor, “Lorentz Group Equivariant Neural Network for Particle Physics,”   (2020), arXiv:2006.04780 [hep-ph] .
  52. Shiqi Gong, Qi Meng, Jue Zhang, Huilin Qu, Congqiao Li, Sitian Qian, Weitao Du, Zhi-Ming Ma,  and Tie-Yan Liu, “An efficient Lorentz equivariant graph neural network for jet tagging,” JHEP 07, 030 (2022), arXiv:2201.08187 [hep-ph] .
  53. Alexander Bogatskiy et al., “Symmetry Group Equivariant Architectures for Physics,” in 2022 Snowmass Summer Study (2022) arXiv:2203.06153 [cs.LG] .
  54. Shikai Qiu, Shuo Han, Xiangyang Ju, Benjamin Nachman,  and Haichen Wang, “A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer,”   (2022), arXiv:2203.05687 [hep-ph] .
  55. Amit Chakraborty, Sung Hak Lim,  and Mihoko M. Nojiri, “Interpretable deep learning for two-prong jet classification with jet spectra,” JHEP 07, 135 (2019), arXiv:1904.02092 [hep-ph] .
  56. Partha Konar, Vishal S. Ngairangbam,  and Michael Spannowsky, “Energy-weighted message passing: an infra-red and collinear safe graph neural network algorithm,” JHEP 02, 060 (2022), arXiv:2109.14636 [hep-ph] .
  57. Alexis Romero, Daniel Whiteson, Michael Fenton, Julian Collado,  and Pierre Baldi, “Safety of Quark/Gluon Jet Classification,”   (2021), arXiv:2103.09103 [hep-ph] .
  58. Oliver Atkinson, Akanksha Bhardwaj, Christoph Englert, Partha Konar, Vishal S. Ngairangbam,  and Michael Spannowsky, “IRC-Safe Graph Autoencoder for Unsupervised Anomaly Detection,” Front. Artif. Intell. 5, 943135 (2022), arXiv:2204.12231 [hep-ph] .
  59. Nimrod Segol and Yaron Lipman, ‘‘On Universal Equivariant Set Networks,”  , 1–11 (2019), arXiv:1910.02421 .
  60. François Chollet et al., “Keras,” https://github.com/fchollet/keras (2015).
  61. Martín Abadi et al., “TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems,”   (2016), arXiv:1603.04467 [cs.DC] .
  62. Sergey Ioffe and Christian Szegedy, “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift,”   (2015), arXiv:1502.03167 [cs.LG] .
  63. Diederik P. Kingma and Jimmy Ba, “Adam: A Method for Stochastic Optimization,”  (2014) arXiv:1412.6980 [cs.LG] .
  64. Sashank J. Reddi, Satyen Kale,  and Sanjiv Kumar, “On the Convergence of Adam and Beyond,”  , 1–23 (2019), arXiv:1904.09237 .
  65. Tianqi Chen and Carlos Guestrin, “XGBoost,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16 (ACM, New York, NY, USA, 2016) pp. 785–794.
  66. J. Bergstra, D. Yamins,  and D. D. Cox, “Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures,” in 30th International Conference on Machine Learning, ICML 2013, PART 1 (2013) pp. 115–123.
  67. James Bergstra, Rémi Bardenet, Yoshua Bengio,  and Balázs Kégl, “Algorithms for hyper-parameter optimization,” in Advances in Neural Information Processing Systems 24: 25th Annual Conference on Neural Information Processing Systems 2011, NIPS 2011 (2011).
Citations (5)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now

Whiteboard

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Generate Now

Continue Learning

We haven't generated follow-up questions for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Generate Now

Collections

Sign up for free to add this paper to one or more collections.

User Circle Single Streamline Icon: https://streamlinehq.com Sign Up