If you're looking to join a Scientific Machine Learning (SciML) Community at Leeds or set up your own at your university, then you've come to the right place. Here’s the story of where the Leeds SciML community came from, who we are and where we’re going.
Where did we come from?
The SciML community at Leeds started during the pandemic with an email I sent to a handful of friends I knew were applying machine learning in their research: “Do you want to have a monthly coffee and chat about ML?”
They all said yes!
The aim was to create an environment where we could all share experiences and help each other in our work.. But also to laugh, discuss and ponder of the crazy, world of AI/ML. I envisioned a place where we could specifically focus on applying machine learning in the physical sciences.
Over the following months more people got entrained: a website was set up (https://sciml-leeds.github.io/) and in the summer Leeds Institute for Data Analytics (LIDA) funded a workshop for us where we brainstormed how to build and grow the community. LIDA brings together applied research groups and data scientists from all disciplines, opening up new opportunities to collaborate on funding bids and team up with industry.
After a fun day of sketching and discussing, we had a plan of what we wanted:
1) to support (read: pay for) training each other in techniques where we already had experience in
2) training by experts outside of Leeds to learn new techniques, through seminars, visits and workshops
3) to create a plan for getting better access to the infrastructure we need for machine learning research
We were given a home in LIDA, a place to meet every first Friday of the month at 11am (with coffee/tea and cake supplied by LIDA too!) This recurrent time and location has been key; it’s a place to meet new people who work in SciML, brainstorm on problem solving collectively and build momentum through continuity.
Since July 2022, we’ve started building, as a community, the things we identified we needed for doing scientific machine learning at Leeds:
- Researcher-lead training materials (Jupyter notebooks) in specific machine learning techniques taught through the researcher’s scientific application: https://github.com/cemac/LIFD_ENV_ML_NOTEBOOKS. Each notebook was created by a researcher in our community in collaboration CEMAC (Leeds Centre for Environmental Modelling And Computation) to demonstrate how they’ve used machine learning in their research, supported by funding from LIFD (Leeds Institute for Fluid Dynamics)
- Community organised seminars given by external researchers who are at the cutting-edge of SciML (recordings are available on our website)
- Inventory of GPU infrastructure available for researchers applying machine learning at Leeds
- Hands-on-SciML sessions lead by individuals in our community: Based on the monthly meet-up chats and discussions we ask someone applying a technique that the community would like to learn to run a Hands-on-SciML session. Through generous funding from LIFD, we can pay ourselves to become the teacher who prepares and runs a two-hour session on a specific machine learning technique taught through their own work. In these a 1hr intro talk with discussion is followed by cake/coffee/tea and then 1hr hands-on coding session where we attempt to solve challenges set by the teacher.
- Monthly meet-ups at fixed time and place: first Friday of every month Worsley Building, 11th Floor, room 11.87.
So, how do you get going setting up at SciML community at your organisation?
It’s easy! Just seek out a few researchers around you who are interested in applying machine learning in their own research and find a monthly meet-up time that works for you. If you follow our simple recipe for the meet-up structure above, you’ll find your community quickly will grow! It is a good idea to set up a website too so the community members can easily find people that overlap in application domain/machine learning technique (you’re welcome to fork our website, the code is on github). If you have questions about getting started or want to set up joint seminars, training etc. with us in Leeds, please get in touch!
The SciML community is for anyone at Leeds who is interested in applying machine learning in their research – this includes you! Please come along to our next monthly meetup (first Friday of every month Worsley Building, 11th Floor, room 11.87) and add yourself to our website - everyone is welcome 🙂
We’re a diverse group of individuals doing research across the physical sciences studying for example iceberg dynamics, air-quality prediction, cardiovascular flow, masonry lined tunnel joints, computer-aided design, detection/location of earthquakes/icequakes, crop yield and resilience, aerosol interactions and climate-impacts of convective clouds, groundwater flows and so much more (please see our website for more). To make predictions and understand processes in these physical applications we are utilising both supervised, self-supervised and un-supervised learning, applying techniques such as deep convolutional networks, generative neural networks, Gaussian processes, reservoir computing, SINDy, physics-informed neural networks (PINN) and random forests. We have active collaboration with people developing machine learning techniques in the Department of Computer Science at Leeds, nationally through the Alan Turing Institute and British Antarctic Survey and internationally through the Frontiers Development Lab and European Space Agency.
We create and run our own training sessions (Hands-on-SciML), produce our own training materials (in collaboration with other organisations in Leeds, e.g. CEMAC and LIFD) typically in the form of Jupyter notebooks (e.g. https://cemac.github.io/LIFD_ENV_ML_NOTEBOOKS/), we invite external researchers at the cutting-edge of SciML to give seminars (if you’d like to give a talk, please get in touch!) and most importantly we are a place to seek out help and provide help to each other. If you’re curious about any of these things come along to our next monthly meet-up!
Where are we going?
Over the next year we will be continuing our monthly meet-ups, external seminars and Hands-on-SciML session, and this summer we’ll be organising our first hackathons (supported by LIFD and CEMAC). We are setting up travel grants for community members to gain external scientific machine learning knowledge. We aim to provide a bottom-up conduit for allowing researchers at Leeds to inform how scientific machine learning is supported and applied at University of Leeds, giving clear direction to where we think the university could go in this exciting journey. Through the training materials and sessions we do, we hope to build the foundation for applying machine learning in research at Leeds and supporting this effort internationally.
And of course, we hope to see you joining our community!
Leif and the rest of the SciML Community Committee (Eilish O’Grady, Maria Taccari, Naomi Shakespeare-Rees and Sam Bancroft)