The recently formed LIDA Environment theme has started fostering cross-university collaboration between researchers who apply machine learning in their day-to-day research in the physical sciences at the University of Leeds. This has crystallised around the Scientific Machine Learning (SciML) Community, connecting early-career machine learning practitioners.
Over the past year, the SciML Community has hosted numerous events:
- Monthly informal ML meetups on the first Friday of each month at 11am, open to all researchers.
- Hybrid seminars focused on machine learning applications in climate, point cloud processing, explainable AI, and more. External speakers have included researchers from eg. Imperial, MIT, Princeton.
- PhD student-led workshops on physics-informed neural networks and reinforcement-learning.(Future plans include workshops on equifinality in different models and semi-supervised learning.)
- A summer hackathon where 30 researchers dived into the exciting world of machine learning applied to environmental science. In this one-day event, the participants worked with satellite imagery provided by the British Antarctic Survey to identify the ice-water interface of sea ice.
The group has compiled resources to support new ML researchers including GPU access, coding tutorials, and best practices for running models. (https://sciml-leeds.github.io)
Furthermore, it has expanded its open-source training materials on applying ML in the physical sciences (https://cemac.github.io/LIFD_ENV_ML_NOTEBOOKS/).
Moving forward, the SciML Community plans to continue growing its seminar series and workshops. There are goals to increase cross-department collaboration on projects utilising ML for environmental modeling, weather prediction, and other domains.
SciML Catch up
Catch up on all your favourite SciML talks on our YouTube channel, where the team have their own playlist.
With over 12k views, our pick of the year has to be the Physics-informed neural networks talk by Fergus Shone, PhD Researcher at the University of Leeds.
Check out the Environment Scientific Machine Learning playlist