Closing date: 10 July 2021
An exciting opportunity at the cutting edge of mathematical data science, in this fully-funded PhD project you will study and develop novel machine learning algorithms for high-precision causal relationships from observational data. Modern machine learning techniques, such as deep learning, are purely associational, that is, they learn to associate inputs to outputs. In practice, observational data captures multiple relationships between variables other than the intended relationship, and modern, high capacity algorithms such as deep learning are particularly prone to finding such relationships. This is a serious problem for high-stakes applications such as healthcare, where spurious, non-causal predictions can have life-altering consequences. By contrast, appropriate experimental data can eliminate these spurious effects but it is often logistically difficult, if not impossible, to carry out such experiments.
Methods from causal inference based on formalisms such as Pearl’s do-calculus, allow estimates of causal relationships from observational data, but require fully probabilistic modelling of all the variables involved. This is largely incompatible with modern, high-capacity machine learning algorithms which are mostly non-probabilistic and where training is performed using gradient descent. The aim of this research is to develop non-probabilistic machine learning algorithms for causal inference, which combine both the causal correctness of classical causal inference methods, and the high capacity of modern machine learning algorithms. The project will have immediate applications to problems at the leading-edge of digital health, such as wearable devices for monitoring the symptoms of neurological disorders.
About the project
You will join the group headed by Dr Max Little at the Computer Science department of the University of Birmingham. The group combines the mathematical theory of machine learning and signal processing, with applications to digital health. There will be multiple opportunities to interact with the wider community at Birmingham CS and across disciplines. This well-established group has a track record of successful supervision of PhD students from diverse backgrounds, placing them in both industrial and academic positions. The subject of this PhD project falls into the broad area of machine learning, signal processing, probabilistic modelling and causal inference, therefore, a solid background in physical applied mathematics, mathematical statistics and/or physics would be appropriate. Programming skills will be a definite advantage, as it will be necessary to carry out extensive computational experiments.
We want our PhD student cohorts to reflect our diverse society. UoB is therefore committed to widening the diversity of our PhD student cohorts. UoB studentships are open to all and we particularly welcome applications from under-represented groups, including, but not limited to BAME, disabled and neuro-diverse candidates. We also welcome applications for part-time study.
First or Upper Second Class Honours undergraduate degree and/or postgraduate degree with Distinction (or an international equivalent). We also consider applicants from diverse backgrounds that has provided them with equally rich relevant experience and knowledge. Full-time and part-time study modes are available. If your first language is not English and you have not studied in an English-speaking country, you will have to provide an English language qualification.
The position offered is for three and a half years full-time study. The value of the award is stipend; £15,285 pa; tuition fee: £4,407. Awards are usually incremented on 1 October each year.
Call for Events is now open! We're supporting Members and Expert Fellows to lead activities that explore aspects of TIPS in the Digital Economy. We will help to organise the activity with up to £5,000 to cover the associated costs.