Kaggle competition

I was the main organizer of “IceCube – Neutrinos in Deep Ice”, a public machine learning competition hosted on Kaggle in collaboration with the Kaggle team, IceCube, and the Munich Data Science Institute.

Participants were given a dataset of 138 million simulated neutrino interactions of all flavors and energies (100 GeV to 100 PeV) and tasked with reconstructing the direction a neutrino came from, scored by the mean angular opening angle between the true and predicted directions. The competition ran from January to April 2023, with $50k in prize money, over 900 participating teams, and more than 11,000 submissions. The winning solutions reached angular resolutions better than 5° for cascade events above 10 TeV and better than 0.5° for tracks — competitive with or exceeding IceCube’s own state-of-the-art reconstructions at the time.

Further Information

Co-organizers: Sohier Dane, Ashley Chow (Kaggle), Lukas Heinrich (TUM), Rasmus Ørsøe (TUM)

Competition page: https://www.kaggle.com/competitions/icecube-neutrinos-in-deep-ice

• Public Kaggle Competition “IceCube – Neutrinos in Deep Ice”: https://arxiv.org/abs/2307.15289
• IceCube – Neutrinos in Deep Ice: The top 3 solutions from the public Kaggle competition: https://arxiv.org/abs/2310.15674