Robert Rouse(He/Him)University of Cambridge
Dr Robert Edwin Rouse is a researcher in the Department of Computer Science at the University of Cambridge, where he works on the application of mathematical and computational methods, including machine learning, to problems in landscape regeneration, predicting the risk and impact of climate phenomena, including floods and heatwaves, and the decarbonisation of the built environment. He is the recipient of a Fellowship from the Royal Commission for the Exhibition of 1851 and the Hawley Medal from the Worshipful Company of Engineerings. He completed his PhD at the University of Cambridge in 2023 and previously received Master’s degrees from Cambridge, Imperial, and the Royal College of Art. |
 |
TalksGaussian & Neural Processes in Hydrology22-Apr-24Machine learning approaches are rapidly changing how modelling is performed in the environmental sciences, with data driven methodologies being slated for operational use in atmospheric and oceanic forecasting. However, despite the capabilities of machine learning models as empirical models for accurately predicting river system behaviour, they face adoption challenges due to concerns around a perceived lack of interpretability and out-of-sample generalisation. Gaussian Process models can offer practitioners a more interpretable model that generalises well and enables extrapolation beyond the training data, albeit at high computational cost. In this research, we investigate the use of Gaussian Processes alongside the related Sparse Variational Gaussian Processes and hybrid Neural Processes and the advantages and disadvantages of these approaches in applying these models to hydrological problems, in terms of general performance, handling extremes, and computational efficiency. |