Hacking the Universe: Leveraging Data Science in High-Energy Astrophysics and Beyond

Daniela Huppenkothen – U. Amsterdam

In data science, the concept of ‘hacking’ involves ingeniously repurposing solutions from diverse fields to tackle new and complex challenges. My group uses this approach to address fundamental astrophysical questions by drawing on methods and tools originally built for disparate problems in industry or other academic fields. This talk focuses into two such questions, spotlighting the breadth of research areas in astrophysics where data science proves invaluable.

Firstly, I will showcase how we use data-driven discovery to unravel the physics behind magnetars—strongly magnetized neutron stars—responsible for triggering and emitting X-ray and (sometimes) radio bursts. In the spirit of hacking, I will also highlight how we apply some of the same methods to other astronomical fields, from asteroids to accreting supermassive black holes. I will then go on to discuss how my group uses machine learning models to accelerate astrophysical inference in black hole X-ray binaries by several orders of magnitude, and how they enable us to ask new kinds of research questions with existing data sets. Dispelling the misconception that machine learning tools are only applicable to datasets from large surveys, I will show how we deploy them extremely effectively when our data and models are especially complex.

Throughout, I will emphasise the collaborative nature of interdisciplinary research, underscoring the interconnectedness among communities, methodologies, tools and software across domains. Fostering effective and equitable collaborations is a key requirement for successful interdisciplinary research, and I will highlight recent results in leveraging hackathons for successful community building..