About the project
How much should a clinician trust a machine learning prediction, and where does the doubt come from? This project will unite distributional and distribution-free uncertainty quantification into one theory, and turn it into reliable tools for high-stakes clinical decisions, such as diagnosing glaucoma and modelling neurodegeneration.
Uncertainty in a prediction has several sources: limited training data, differences between hospitals and patient groups, and the inherent unpredictability of each outcome. Knowing which source dominates tells a clinician what to do next.
Two traditions study this:
- distributional approaches, such as hierarchical models, capture the full spread of plausible predictions and show where uncertainty comes from
- distribution-free approaches, such as conformal prediction, give guarantees without assuming a model
The two are complementary, yet have largely developed apart. You'll bring them together into one unified body of work.
You will:
- master the theory of uncertainty in machine learning: where it comes from, how it is represented, and what can be formally guaranteed about it
- develop methods that combine hierarchical uncertainty models with conformal guarantees
- establish what can be guaranteed for patients from hospitals seen during training, and from entirely new one
- design decision rules that know when to defer to a clinician, with provable error control
- apply them to problems such as glaucoma screening from retinal images and multiple sclerosis progression from brain magnetic resonance imaging (MRI)
This is foremost a fundamental research project: most of your time will go into theory and new methods for uncertainty quantification, while applications keep the work grounded. Clinicians at University Hospital Southampton, our collaborators, will help you validate and refine the methods.
The project is a joint venture between the Vision, Learning and Control group and the Digital Health and Biomedical Engineering group in the School of Electronics and Computer Science. You'll have access to the University's Iridis supercomputer and a lively machine learning and digital health community.
The School of Electronics and Computer Science is committed to promoting equality, diversity inclusivity as demonstrated by our Athena SWAN award. We welcome all applicants regardless of their gender, ethnicity, disability, sexual orientation or age, and will give full consideration to applicants seeking flexible working patterns and those who have taken a career break. The University has a generous maternity policy, onsite childcare facilities, and offers a range of benefits to help ensure employees’ well-being and work-life balance. The University of Southampton is committed to sustainability and has been awarded the Platinum EcoAward.
