Statistics and Machine Learning (EPSRC Centre for Doctoral Training) DPhil

University of Oxford

The Statistics and Machine Learning CDT is a four-year DPhil research programme (or eight years if studying part-time). It will train the next generation of researchers in statistics and machine learning, who will develop widely-applicable novel methodology and theory and create application-specific methods, leading to breakthroughs in real-world problems in government, medicine, industry and science.

This is the Oxford component of StatML, an EPSRC Centre for Doctoral Training (CDT) in Statistics and Machine Learning, co-hosted by Imperial College London and the University of Oxford. The CDT will provide you with training in both cutting-edge research methodologies and the development of business and transferable skills – essential elements required by employers in industry and business.

You will undertake a significant, challenging and original research project, leading to the award of a DPhil. Given the breadth and depth of the research teams at Imperial College and at the University of Oxford, the proposed projects will range from theoretical to computational and applied aspects of statistics and machine learning, with a large number of projects involving strong methodological/theoretical developments together with a challenging real problem. A significant number of projects will be co-supervised with industry.

You will pursue two mini-projects during your first year (specific timings may vary for part-time students), with the expectation that one of them will lead to your main research project. At the admissions stage you will choose a mini-project. These mini-projects are proposed by the department's supervisory pool and industrial partners. You will be based at the home institution of your main supervisor of the first mini-project.

If your studentship is funded or co-funded by an external partner, the second mini-project will be with the same external partner but will explore a different question.

You will then begin your main DPhil project at the beginning of the third term, which can be based on one of the two mini-projects. Where appropriate for the research, your project will be run jointly with the CDT’s leading industrial partners, and you will have the chance to undertake a placement in data-intensive statistics with some of the strongest statistics groups in the USA, Europe and Asia.

Alongside your research projects you will engage with taught courses each lasting for two weeks. Core topics will be taught during at the beginning of your first year (specific timings may vary for part-time students) and are:

• Modern Statistical Theory
• Statistical Machine Learning;
• Causality; and
• Bayesian methods and computation.

You will also be required to take a number of optional courses throughout the four years of the course, which could be made up of choices from the following list: Bayesian nonparametrics; high-dimensional statistics; advanced optimisation; networks; reinforcement learning; large language models; conformal inference, variational Bayes and advance Bayesian computations, dynamical and graphical modelling of multivariate time series, modelling events; and deep learning. Optional modules last two weeks and are delivered in a similar format to the core modules.