Details of Award
NERC Reference : NE/Y003675/1
Bayesian Machine Learning as a tool for Climate Scientists: An In-Person Workshop at the University of Cambridge
Training Grant Award
- Lead Supervisor:
- Dr H B Moss, University of Cambridge, Applied Maths and Theoretical Physics
- Grant held at:
- University of Cambridge, Applied Maths and Theoretical Physics
- Science Area:
- Atmospheric
- Earth
- Freshwater
- Marine
- Terrestrial
- Overall Classification:
- Atmospheric
- ENRIs:
- Environmental Risks and Hazards
- Global Change
- Science Topics:
- Climate & Climate Change
- Artificial Intelligence
- Machine Learning (AI)
- Statistics & Appl. Probability
- Applied Statistics
- Bayesian Methods
- Curve Fitting
- Data Exploration
- Design of Experiments
- Kernel Mathods
- Spatial Statistics
- Statistical Uncertainty
- Abstract:
- Background: To gain a better understanding of our climate, scientists analyse large amounts of complex data collected across various regions and ecosystems worldwide. Consequently, Machine Learning (ML) has proven to be an invaluable tool for the modern climate science researcher, as it enables more effective analysis and interpretation of these vast quantities of data. However, given the large array of techniques available from statistics, data science, and computer science, there is a need for specialised training to ensure their impact is maximised. Impact: We propose a dynamic, practical programme that leverages the UK's excellence in ML research to equip the next generation of NERC-funded climate scientists with the necessary knowledge, skills, and support to integrate ML into their research during their PhDs and future careers. Our course, designed for 30 participants, provides a national resource by offering highly relevant ML expertise to UK environmental scientists at an appropriate point in their careers, just as ML is becoming a growing part of world-leading environmental science. The programme, run by the Institute of Computing for Climate Science (ICCS) in Cambridge, includes hands-on exercises that use real-world climate data, providing participants with practical experience in a collaborative, project-based environment, making them highly employable in both academia and industry. Uniqueness: We believe our proposed training course is the first of its kind for NERC PhD students in the UK and, as such, has two nonstandard foci that are critical when applying ML to climate science but often overlooked in standard introductory courses. By focusing on these topics, we will prepare students to integrate ML into their research effectively and communicate their findings to policymakers. Firstly, to ensure ML is appropriately deployed, our programme will focus heavily on sharing knowledge of its limitations. As a field enjoying significant publicity, ML is often oversold to practitioners and there is a disconnect between typical ML advancements and their ability to solve real-world scientific problems. Therefore, in order to successfully use ML to support their research, students must learn how to appropriately interrogate and interpret new ML advances. Secondly, as understanding uncertainty is crucial when communicating potential risks and impacts of climate change, we will focus on Bayesian ML, which, unlike many traditional ML methods, can quantify uncertainty in interpretations and predictions. We will leverage the substantial Bayesian ML expertise in Cambridge and the active community of researchers across the university and the British Antarctic Survey already applying these methods to problems in climate science. Tutorials will be provided on corel topics like Bayesian neural networks and Gaussian processes. Provisional schedule: Our 3-day in-person programme is designed to be highly interactive. Participants will have opportunities to network, share ideas, learn from the workshop organisers and each other, and participate in professional development training. Day 1: What does it mean to be Bayesian? Bayesian neural networks A hands-on Introduction to Gaussian processes with JAX Professional development: generating impact (EW Group, a leading global DEI consultancy) Poster session Day 2: How to build and debug ML pipelines Short talks on existing climate science ML projects (multidisciplinary speakers) Lab session: Analysing an ocean dynamics dataset Reproducible research/ open science Professional development: DEI training (EW Group) Day 3: Team hackathon to tackle a climate science challenge Open office hour for students' problems (Research Software Team, ICCS)
- NERC Reference:
- NE/Y003675/1
- Grant Stage:
- Completed
- Scheme:
- Doctoral Training
- Grant Status:
- Closed
- Programme:
- Advanced Training
This training grant award has a total value of £41,412
FDAB - Financial Details (Award breakdown by headings)
| Total - Other Costs |
|---|
| £41,412 |
If you need further help, please read the user guide.