Deep Probabilistic Machine Learning

This project will develop scalable approaches to deep non-parametric probabilistic models that use approximate inference techniques to learn the structure of the model. The project requires that the development of practical, interpretable models, with latent variables that can be used by clinicians and non-academics in a meaningful way. We also aim to build distributed user-centric data models, in which the learning occurs across distributed devices, through the paradigm of differential privacy. The successful candidate will be able to demonstrate knowledge of a wide range of machine learning techniques (in particular probabilistic modelling) and practical experience handling data which is noisy, sparse and/or of high dimensionality.

Development and application of machine learning techniques for characterisation and quantification of change in time-lapse resistivity monitoring

This project was part of a collaborative effort and joint supervision with British Geological Survey addressing the need for automatic feature detection and interpretation 3D time-lapse images of subsurface geology obtained through electrical resistivity tomography. The interdisciplinary work will combine new techniques in computer vision and change recognition in increased dimension to identify, predict and understand changes to and caused by complex hydro-geophysical processes. Results were be verified through a combination of synthetic, experimental and field models.