Loading [MathJax]/jax/output/HTML-CSS/jax.js

IceCube – Neutrinos in Deep Ice

Habib Bukhari (Ashburn, VA, USA) ; Dipam Chakraborty (Bangalore, India) ; Philipp Eller (Physics Department, TUM School of Natural Sciences, Technical University of Munich, Garching, 85748, Germany; Munich Data Science Institute, Technical University of Munich, Garching, 85748, Germany) ; Takuya Ito (Tokyo, Japan) ; Maxim Shugaev (Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA, 22904-4745, USA) ; et al. - Show all 6 authors

During the public Kaggle competition “IceCube – Neutrinos in Deep Ice”, thousands of reconstruction algorithms were created and submitted, aiming to estimate the direction of neutrino events recorded by the IceCube detector. Here we describe in detail the three ultimate best, award-winning solutions. The data handling, architecture, and training process of each of these machine learning models is laid out, followed up by an in-depth comparison of the performance on the Kaggle datatset. We show that on cascade events in IceCube above 10 TeV, the best Kaggle solution is able to achieve an angular resolution of better than 5 , and for tracks correspondingly better than 0.5 . These results indicate that the Kaggle solutions perform at a level comparable to the current state-of-the-art in the field, and that they may even be able to outperform existing reconstruction resolutions for certain types of events.

{
  "_oai": {
    "updated": "2024-07-19T00:31:40Z", 
    "id": "oai:repo.scoap3.org:86560", 
    "sets": [
      "EPJC"
    ]
  }, 
  "authors": [
    {
      "affiliations": [
        {
          "country": "USA", 
          "value": "Ashburn, VA, USA"
        }
      ], 
      "surname": "Bukhari", 
      "given_names": "Habib", 
      "full_name": "Bukhari, Habib"
    }, 
    {
      "affiliations": [
        {
          "country": "India", 
          "value": "Bangalore, India"
        }
      ], 
      "surname": "Chakraborty", 
      "given_names": "Dipam", 
      "full_name": "Chakraborty, Dipam"
    }, 
    {
      "affiliations": [
        {
          "country": "Germany", 
          "value": "Physics Department, TUM School of Natural Sciences, Technical University of Munich, Garching, 85748, Germany", 
          "organization": "Technical University of Munich"
        }, 
        {
          "country": "Germany", 
          "value": "Munich Data Science Institute, Technical University of Munich, Garching, 85748, Germany", 
          "organization": "Technical University of Munich"
        }
      ], 
      "surname": "Eller", 
      "email": "philipp.eller@tum.de", 
      "full_name": "Eller, Philipp", 
      "given_names": "Philipp"
    }, 
    {
      "affiliations": [
        {
          "country": "Japan", 
          "value": "Tokyo, Japan"
        }
      ], 
      "surname": "Ito", 
      "given_names": "Takuya", 
      "full_name": "Ito, Takuya"
    }, 
    {
      "affiliations": [
        {
          "country": "USA", 
          "value": "Department of Materials Science and Engineering, University of Virginia, Charlottesville, VA, 22904-4745, USA", 
          "organization": "University of Virginia"
        }
      ], 
      "surname": "Shugaev", 
      "given_names": "Maxim", 
      "full_name": "Shugaev, Maxim"
    }, 
    {
      "affiliations": [
        {
          "country": "Germany", 
          "value": "Physics Department, TUM School of Natural Sciences, Technical University of Munich, Garching, 85748, Germany", 
          "organization": "Technical University of Munich"
        }, 
        {
          "country": "Germany", 
          "value": "Munich Data Science Institute, Technical University of Munich, Garching, 85748, Germany", 
          "organization": "Technical University of Munich"
        }
      ], 
      "surname": "\u00d8rs\u00f8e", 
      "email": "rasmus.orsoe@tum.de", 
      "full_name": "\u00d8rs\u00f8e, Rasmus", 
      "given_names": "Rasmus"
    }
  ], 
  "titles": [
    {
      "source": "Springer", 
      "title": "IceCube \u2013 Neutrinos in Deep Ice"
    }
  ], 
  "dois": [
    {
      "value": "10.1140/epjc/s10052-024-12977-2"
    }
  ], 
  "publication_info": [
    {
      "page_end": "19", 
      "journal_title": "European Physical Journal C", 
      "material": "article", 
      "journal_volume": "84", 
      "artid": "s10052-024-12977-2", 
      "year": 2024, 
      "page_start": "1", 
      "journal_issue": "6"
    }
  ], 
  "$schema": "http://repo.scoap3.org/schemas/hep.json", 
  "acquisition_source": {
    "date": "2024-07-19T00:31:23.307288", 
    "source": "Springer", 
    "method": "Springer", 
    "submission_number": "08538b04456611ef9fecbe60ec5a7b90"
  }, 
  "page_nr": [
    19
  ], 
  "license": [
    {
      "url": "https://creativecommons.org/licenses//by/4.0", 
      "license": "CC-BY-4.0"
    }
  ], 
  "copyright": [
    {
      "holder": "The Author(s)", 
      "year": "2024"
    }
  ], 
  "control_number": "86560", 
  "record_creation_date": "2024-06-25T15:30:23.404314", 
  "_files": [
    {
      "checksum": "md5:858c8d73d83a6dfdf36cf614b7b8af85", 
      "filetype": "xml", 
      "bucket": "554930a7-cd9e-4f4c-8d4a-01edd5f6f5ba", 
      "version_id": "9acd0cd6-487c-4b28-a3a3-a8c812eb6138", 
      "key": "10.1140/epjc/s10052-024-12977-2.xml", 
      "size": 15992
    }, 
    {
      "checksum": "md5:8d685682ebbb3f5a49fb48fd3805c9a7", 
      "filetype": "pdf/a", 
      "bucket": "554930a7-cd9e-4f4c-8d4a-01edd5f6f5ba", 
      "version_id": "5bf7bcbc-745d-4996-8e98-6384df4b69d8", 
      "key": "10.1140/epjc/s10052-024-12977-2_a.pdf", 
      "size": 2645364
    }
  ], 
  "collections": [
    {
      "primary": "European Physical Journal C"
    }
  ], 
  "arxiv_eprints": [
    {
      "categories": [
        "astro-ph.HE", 
        "hep-ex", 
        "physics.data-an"
      ], 
      "value": "2310.15674"
    }
  ], 
  "abstracts": [
    {
      "source": "Springer", 
      "value": "During the public Kaggle competition \u201cIceCube \u2013 Neutrinos in Deep Ice\u201d, thousands of reconstruction algorithms were created and submitted, aiming to estimate the direction of neutrino events recorded by the IceCube detector. Here we describe in detail the three ultimate best, award-winning solutions. The data handling, architecture, and training process of each of these machine learning models is laid out, followed up by an in-depth comparison of the performance on the Kaggle datatset. We show that on cascade events in IceCube above 10 TeV, the best Kaggle solution is able to achieve an angular resolution of better than 5 $$^{\\circ }$$  <math> <mmultiscripts> <mrow></mrow> <mrow></mrow> <mo>\u2218</mo> </mmultiscripts> </math>  , and for tracks correspondingly better than 0.5 $$^{\\circ }$$  <math> <mmultiscripts> <mrow></mrow> <mrow></mrow> <mo>\u2218</mo> </mmultiscripts> </math>  . These results indicate that the Kaggle solutions perform at a level comparable to the current state-of-the-art in the field, and that they may even be able to outperform existing reconstruction resolutions for certain types of events."
    }
  ], 
  "imprints": [
    {
      "date": "2024-06-25", 
      "publisher": "Springer"
    }
  ]
}
Published on:
25 June 2024
Publisher:
Springer
Published in:
European Physical Journal C , Volume 84 (2024)
Issue 6
Pages 1-19
DOI:
https://doi.org/10.1140/epjc/s10052-024-12977-2
arXiv:
2310.15674
Copyrights:
The Author(s)
Licence:
CC-BY-4.0

Fulltext files: