What is it about?
This is the fourth workshop dedicated to GraphNeT – A deep learning library for neutrino telescopes. The goal is to bring together researchers working at the intersection of neutrino telescope experiments and machine learning to meet likeminded researchers, discuss the latest progress, and develop new solutions to physics challenges by applying deep learning where it matters, using GraphNeT.
This May, we're celebrating the launch of GraphNeT 2.0 which extends functionality beyond graph neural networks into other deep learning paradigms like normalizing flows, transformers, autoencoders and several of the winning solutions from the IceCube Kaggle Competition.
The workshop, and particularly the two half-day hackathons, are focussed on putting the GraphNeT framework into use for physics — in IceCube, P-ONE, KM3NeT and other experiments. In this way, we hope to foster collaboration on common tools in order to advance physics research faster than individual experiments can on their own.
During the workshop, participants will be presented an opportunity to get involved in a joint publication that aims to provide the first-ever apples-to-apples comparison of deep learning techniques from different neutrino telescopes on a series of open-source datasets specfically prepared for this workshop by the team behind open-source simulation tool Prometheus.
Who is it for?
The GraphNeT user community can be broken into two broad categories:
- a) Users with a high level of deep learning expertise that translate new, novel techniques from deep learning research into solutions within the context of neutrino telescopes.
- b) Neutrino physicists that are domain experts and require sharper methods to carry out their physics analyses.
Physics analyses often require expert knowledge in both of these categories, but very few people are experts in both. Therefore, a key goal of GraphNeT is to provide a common framework for both types of users and to foster partnerships across these two disciplines and across experiments. The end-goal is to open doors that are currently closed; to turn evidence into discoveries and help find evidence where there currently is none.
While the workshop's presentations can feature most work at the intersection of ML and neutrino telescopes, e.g. new technical developments or physics challenges, hackathon participants ought to have an interest in trying out the GraphNeT code — either because they’re open to using it as their main framework, or because they want to see how other people are working with deep learning.
Where is it held?
The workshop will be held in Munich, the capitol of the German state of Bavaria. Specifically, the workshop activities will take place at the TUM Campus in Garching, slightly north of Munich, at the TUM Institute for Advanced Study (IAS).
There may be some limitations to the workshop's capacity, in which case preference will be given to attendees who fit well in to either of the above categories of GraphNeT users.