Bio
Ollie is working on the Lloyds Registry Foundation funded London air quality project. Ollie was previously a student at the University of Warwick where he got a first class MEng in Computer Science. His 4th year thesis was on modelling air pollution in London. Ollie also spent three summers as a Software Engineer Intern at Modality Systems in which he was creating prototypes using new technologies, worked in testing and worked on their main products.
Research interests
Throughout their time at Turing, Ollie has published 4 papers at top machine learning conferences. In Ollie's doctoral research, they have focused on Bayesian methodologies to incorporate inductive biases and prior constraints in spatio-temporal models, while ensuring computational scalability. To achieve this, they have developed Gaussian process-based models that adhere to constraints such as positivity and boundedness, as well as those inspired by physics, and can accommodate observations from diverse sources. Throughout, they have focused on computational efficiency and have introduced a variational framework that enables these models to scale linearly in the temporal dimension by leveraging Kalman smoothing algorithms.
Selected publications and papers
‘Spatio-temporal variational Gaussian processes’: Oliver Hamelijnck*, William J. Wilkinson*, Niki A. Loppi, Arno Solin, Theodoros Damoulas, Conference on Neural Information Processing Systems (NeurIPs), 2021
‘Transforming Gaussian processes with normalizing flows’: Juan Maronãs*, Oliver Hamelijnck*, Jeremias Knoblauch, Theodoros Damoulas, International Conference on Artificial Intelligence and Statistics (AISTATS) 2021
‘Non-separable Non-stationary random fields’: Kangrui Wang, Oliver Hamelijnck, Theodoros Damoulas, Mark Steel’, The International Conference on Machine Learning, 2020
‘Multi-resolution Multi-task Gaussian Processes’, Oliver Hamelijnck and Theodoros Damoulas and Kangrui Wang and Mark Girolami’, Conference on Neural Information Processing Systems (NeurIPs), 2019
