Dr Aiden Durrant

Biography

I am primarily interested in how to learn high-quality image representations and importantly how to structure such representations so that they are useful for a variety of tasks. Specifically, I am focused on learning representations without human-annotated labels, instead constructing methods to learn concepts from the data itself (Self-Supervised Learning). In addition, how to best structure the knowledge captured by representations is of primary interest, which until recently has been traditionally overlooked when designing these systems. To address this, I am exploring the geometric structure of knowledge and semantics, and developing Machine Learning approaches that directly operate in such geometries to preserve information.

I obtained my BSc (Hons) and MPhil degree in Computer Science at the University of Lincoln, particularly focusing on computer vision and Machine Learning researching machine learning in the setting of nuclear reactor anomaly detection as part of the CORTEX Horizon 2020 project. Directly leading from this work, I completed my PhD in Computing Science at the University of Aberdeen supervised by Professor Georgios Leontidis, and Dr. Mingjun Zhong. Alongside my PhD studies, I have also held positions as an Early Career Researcher at the University of Glasgow and at the University of Aberdeen on a variety of projects applying to Machine Learning to key industrial settings such as Agriculture, Environmental Sustainability and Healthcare.

Externally, I am the Deputy Director of the SICSA Graduate Academy assisting with the coordination of research training and activities across all 14 Scottish Computer Science departments.

External Memberships

Deputy Director of SICSA Student Graduate Academy - Scottish Informatics and Computer Science Alliance 

BMVA 2025 Summer School Co-Organiser

Prizes and Awards

PhD Poster Award, FISA - 2019:

9th European Commission conference on Euratom research and training in safety of reactor systems

Research Overview

My work is primarily focused on unsupervised representation learning and self-supervised learning, with a key interest in the design of architectural and objective functions for structuring the learnt embeddings. Although my work resides in computer vision, I am also interested in the applicability of such methods to all modalities and the intersection of multi-modal learning. 

From a more general perspective, I am also researching alternative geometric manifolds for machine learning models which better capture the underlying structures/priors of the knowledge we intend to capture. Specifically, I am currently focussing on Hyperbolic Deep Learning for embedding semantic hierarchies. 

Regarding application areas, I have a keen interest in environmental monitoring and agriculture, with precedence placed on computer vision.

Funding and Grants

[2024] (Co-PI) £3,500 – Aberdeen Internal Pump-Priming: Comprehensive mapping of glacier discharge using 4D spatio-temporal generative AI modelling.

[2024] (PI) SUSTAIN CDT Studentship + £70,000 In-Kind – Machine Learning Crop Breeding in Precision-Controlled Vertical Farming for a Sustainable Future. 

Teaching Responsibilities

South China Normal University Joint Institute

Module Coordinator - Machine Learning (BSc)

Previously

University of Aberdeen

Co-Module Coordinator for CS4040 - Research Methods 2022-2023 (BSc)

Co-Module Coordinator for CS5062 - Machine Learning 2021-2022 (MSc)

Co-Module Coordinator for CS551G - Data Mining and Visualisation 2020-2021 (MSc)