Who we are

Neutrino telescope experiments around the world produce vast amounts of data every day, with many complex tasks suitable for the use of machine learning (ML) solutions. In addition, they generally share a lot of the characteristics (e.g., non-trivial detector geometries, detector heterogeneity, and physical properties of the medium in which they are embedded) that makes it challenging to employ ML in these expriements.

The Global Neutrino Network (GNN) Machine Learning (ML) Working Group (WG) aims to facilitate the advancement of, and collaboration around, ML methods at neutrino telescope experiments, by providing a forum for openly sharing and discussing results and ideas within ML as it relates to the unique priorities and challenges of neutrino physics.

What we do

The WG aims to organise regular (approx. monthly), online, seminar-style meetings focused on core ML concepts of importance to the community and novel developments at the intersection of ML and neutrino physics. In addition, the WG aims to organise in-person workshops (approx. annually) to allow community members to meet, showcase their work, discuss topics of common interest, and participate in tutorials to learn the core skills to allow them to leverage ML towards their physics goals.

The WG initially intends to focus on:

  • ML modelling tasks, specifically which architectures and techniques works best for, and has the greatest impact on, physics analyses.
  • Pursuing, establishing open/public datasets that can be used for ML-in-physics research.
  • Encouraging, facilitating standardisation of data formats, tools, etc.
  • Encouraging, facilitating open-sourcing ML models and pipelines, e.g. through the GNN ML WG GitHub organisation (see Home or below).
  • Limited, topical discussion of more specific areas (hardware, tools, etc.).

Coordinators