Dr Michael Smith
MScs, PhD
School of Computer Science
Senior Lecturer
Outreach Lead
Member of the Machine Learning research group
Full contact details
School of Computer Science
Regent Court (CS)
211 Portobello
Sheffield
S1 4DP
- Profile
-
Dr Michael Smith currently works at the intersection of (flying) insect behaviour ecology and computer science, in particular on the development of methods to track their location and behaviour in the landscape.
This depends on the development of tags and machine learning techniques for efficient sampling and handling of the sparse data the tags collect, as the bee flys across the landscape.
Previous work focused on the probabilistically handling calibration of air pollution sensors using mobile PM2.5 sensors (transported by motorbike taxis in Kampala), and the problem of source-inference in the resulting datasets. He also worked in the field of Differential Privacy and its applications to Gaussian process (GP) regression and classification and developed an approach to bound all future attacks on GP classifiers within the framework of adversarial examples.
He studied his undergraduate in Computer Science at Warwick university, then in Edinburgh completed his masters in Informatics and Neuroinformatics and a PhD in computational neuroscience, looking at where self-motion cues are processed and integrating, in the human brain (using fMRI). He next went to Kampala (Uganda) to lecture (in 2014) teaching AI to students at Makerere, and is now a senior lecturer at the University of Sheffield in the School of Computer Science in the Machine Learning group.
- Research interests
-
- Gaussian Processes
- Air pollution
- Differential Privacy
- Machine Learning for International Development
- Bumblebee tracking
- Adversarial Examples/bounds using Gaussian Processes
- Publications
-
Journal articles
- Modelling the effects of vegetation and urban form on air quality in real urban environments: A systematic review of measurements, methods, and predictions. Urban Forestry & Urban Greening, 105. View this article in WRRO
- Bumblebee nest departures under low light conditions at sunrise and sunset. Biology Letters, 20(4). View this article in WRRO
- Modelling calibration uncertainty in networks of environmental sensors. Journal of the Royal Statistical Society Series C: Applied Statistics, 72(5), 1187-1209. View this article in WRRO
- Adversarial vulnerability bounds for Gaussian process classification. Machine Learning, 112(3), 971-1009. View this article in WRRO
- A method for low‐cost, low‐impact insect tracking using retroreflective tags. Methods in Ecology and Evolution, 12(11), 2184-2195.
- Differentially private regression and classification with sparse Gaussian processes. Journal of Machine Learning Research, 22. View this article in WRRO
- Hospitalization and mortality following non-attendance for hemodialysis according to dialysis day of the week : a European cohort study. BMC Nephrology, 21(1). View this article in WRRO
- Differentially Private Gaussian Processes.. CoRR, abs/1606.00720.
- The Postsubiculum and Spatial Learning: The Role of Postsubicular Synaptic Activity and Synaptic Plasticity in Hippocampal Place Cell, Object, and Object-Location Memory. The Journal of Neuroscience, 33(16), 6928-6943.
- Fluctuations in the open time of synaptic channels: An application to noise analysis based on charge. Journal of Neuroscience Methods, 210(1), 15-21.
- Physiological signal variability in hMT+ reflects performance on a direction discrimination task. Frontiers in Psychology, 2.
Book chapters
- Modeling the Effects of Vegetation on Urban Air Quality Based on Multiple Environmental Data Sources, Advances in Science, Technology & Innovation (pp. 55-67). Springer Nature Switzerland
Conference proceedings
- Nonparametric gaussian process covariances via multidimensional convolutions. Proceedings of Machine Learning Research, Vol. 206 (pp 8279-8293). Palau de Congressos, Valencia, Spain, 25 April 2023 - 25 April 2023. View this article in WRRO
- Adjoint-aided inference of Gaussian process driven differential equations. Advances in Neural Information Processing Systems (NeurIPS 2022), Vol. 35. New Orleans, LA, USA, 28 November 2022 - 28 November 2022. View this article in WRRO
- Killing four birds with one Gaussian Process: The relation between different test-time attacks. 2020 25th International Conference on Pattern Recognition (ICPR) Proceedings (pp 4696-4703). MIlan, Italy, 10 January 2021 - 10 January 2021. View this article in WRRO
- Multi-task Learning for aggregated data using Gaussian processes. Advances in Neural Information Processing Systems 32 (NeurIPS 2019), Vol. 32 (pp 15050-15060). Vancouver, Canada, 8 December 2019 - 8 December 2019. View this article in WRRO
- Multi-task Learning for aggregated data using Gaussian processes. Proceedings of the conference on Advances in Neural Information Processing Systems (NIPS 2019), Vol. 32. Vancouver, Canada, 8 December 2019 - 8 December 2019. View this article in WRRO
- Differentially private regression with Gaussian processes. Proceedings of the 21st International Conference on Artificial Intelligence and Statistics(84) (pp 1195-1203). Lanzarote, Canary Islands, 9 April 2018 - 9 April 2018. View this article in WRRO
Preprints
- Learning Nonparametric Volterra Kernels with Gaussian Processes, arXiv.
- The Limitations of Model Uncertainty in Adversarial Settings.
- Differentially Private Regression and Classification with Sparse Gaussian Processes, arXiv.
- Multi-task Learning for Aggregated Data using Gaussian Processes, arXiv.
- Gaussian Process Regression for Binned Data, arXiv.
- Modelling the effects of vegetation and urban form on air quality in real urban environments: A systematic review of measurements, methods, and predictions. Urban Forestry & Urban Greening, 105. View this article in WRRO
- Grants
-
- Tracking flying insects: Deploying novel technology to explore the lives of bees, Leverhulme Trust, 02/2026 - 08/2029, £263,018, as PI
- BLE Bee Tracking System, Eva Crane Trust, 05/2025 - 05/2027, £19,902, as Co-I
- Using Data Driven Artificial Intelligence to Reveal Subtle Pesticide Induced Changes in Pollinator Behaviour, Biotechnology and Biological Sciences Research Council, 02/2024 - 10/2025, £249,159, as PI
- AirQo, Industrial, 08/2019 - 07/2023, £197,726, as PI
- Improved Retroreflector Based Tracking for Bees, Eva Crane Trust, 03/2021 - 03/2023, £13,909, as PI
- Foraging distances and nest locations of bumblebees Bombus, Eva Crane Trust, 04/2019 - 12/2020, £5,341, as PI