School of Computing and Mathematical Sciences

Distributed and high-performance AI systems and digital twins

The group conducts multidisciplinary research in machine learning, artificial intelligence (AI), and high-performance solutions for their deployment using:

Self-learning digital twins

A digital twin refers to a representation of a physical object or system. Digital twins can be augmented with a self-learning approach, to permit real-time learning given data produced by its real-world counterpart. A self-learning digital twin can thus mirror the behaviour and the dynamics of its physical counterparts and is enabled to improve its accuracy and predictive capabilities over time. This is achieved through the integration of machine learning algorithms and other AI techniques.

Physics-informed neural networks

Linking modelling and AI, physics-informed neural networks enhance machine learning by integrating domain-specific knowledge from subject matter specialists about the physical or biological object represented by an AI. This is especially useful when data is noisy or incomplete.

Federated learning

It is a decentralised approach to machine learning and AI, where heavy computations are performed in multiple devices (and different physical locations) without exchanging data. In health sciences, it enables collaborative research by allowing institutions to develop predictive models without sharing sensitive patient data.

Group members

Back to top