Many fields make use of large-scale scientific codes — numerical weather and climate prediction, atomic and molecular modelling, and plasma physics to name a few. Developments in machine learning have brought opportunities to advance these models through a "hybrid-modelling" approach. For example, using emulators and leveraging data-driven approaches. To achieve this in these large codebases presents a significant challenge on scientific, software, and computational fronts, however.
This two-day workshop held in Cambridge, seeks to bring together researchers, research software engineers, and modelling centres currently working to tackle these challenges to share their recent advances and expertise. We will hear from a range of people about how they have done so, the approaches and tools used, and what they have learnt. Discussion sessions will explore the key current challenges in hybrid modelling looking to establish best practices.
Sign-up and more details here: https://cambridge-iccs.github.io/ml-coupling-workshop/