Applied Machine Learning
We develop novel machine learning algorithms designed for specific applications.
In Biology, Medicine and Pharmacology, our goal is translational research through approaches that
integrate molecular and systems data. We have recently developed the first method for the
prediction of disease genes for orphan diseases, and the first method for the prediction of the
frequency of drug side effects.
Using a new abstract model of evolution, we are developing new types of genetic algorithm that
can be analysed rigorously as statistical sampling methods, and we seek applications of these.
We are using machine learning in novel ways to devise solutions to a variety of other problems,
including smart-metering protocols that preserve individual privacy, on-line discussion systems
that allow more informative views of large conversations, and faster methods of finding bugs
in video games.
The group also consider the importance of data itself and is interested in the impact of the FAIR
data principles and Open Science in making research more effective.
Natural Language Processing and Understanding
We work on theories and algorithms that allow humans and computers to process, generate and
understand natural languages. Our work ranges from natural language semantics (e.g., formal
semantics in modern type theories) to the development of innovative deep learning approaches
to Natural Language Processing (NLP). Techniques based on our research have been applied to
multiple NLP tasks, including natural language inference, document summarisation, machine
translation, and NLP evaluation. We have organised series of talks, lectures, and workshops
to advance research and promote knowledge sharing in the NLP community.
Safe and Autonomous Intelligent Systems
We work on designing and building intelligent and autonomous systems. Our
research covers a broad range of AI capabilities including planning,
optimisation, negotiation, reasoning, interpretability, safety and security for
multi-agent/robotic applications, cyber-physical systems, and edge computing. We
are active both in theoretical and applied research in these broad areas. Our
recent work has studied specialised topics such as sub-modularity, translation
between different planning formalisms, adversarial robustness, autonomous agent
models and formal verification and synthesis.
The outcomes of our research have been published in top-ranked journals such as
Artificial Intelligence Journal, Journal of Artificial Intelligence Research,
ACM Transactions on Cyber-Physical Systems, ACM Transactions on Computer-Human Interaction,
IEEE Robotics and Automation Letters, and conferences such as AAAI, IJCAI, IROS, ICAPS,
IEEE/RSJ, UAI, ICCPS, CAV and VLDB. These outcomes have been produced in
challenge-driven projects that we lead when applying AI in robotics for extreme
environments, net-zero oceanographic missions, education, healthcare and medical
devices. These projects have been funded by industry and agencies such as the EPSRC,
Innovate UK, the Leverhulme Trust, the National Cyber Security Centre and the EU.
