PhD Studentship: Adapting Large Pretrained Models to Non-Standard Conditions

Summary

Modern AI systems, powered by large pretrained models, have demonstrated exceptional performance across a variety of applications, particularly when trained and deployed in standard conditions. However, real-world scenarios are often far from ideal, presenting challenges such as varying lighting, environmental noise, sensor degradation, and domain shifts. The ability to adapt pretrained models for robust operation in such non-standard conditions is becoming increasingly critical, especially in domains like autonomous systems and disaster response.

Pretrained models are typically developed using datasets that represent standard conditions, which limits their effectiveness in non-standard environments. Direct training for every possible scenario is computationally prohibitive, and the scarcity of labelled data in non-standard conditions exacerbates this issue. These limitations highlight the need for innovative approaches to fine-tune or adapt pretrained models for robust performance in diverse and challenging settings.

This project aims to develop advanced methodologies to adapt pretrained models for non-standard conditions, focusing on strategies such as fine-tuning with limited or weakly labelled data, leveraging domain adaptation and transfer learning to bridge the gap between standard datasets and real-world scenarios, and enhancing robust feature representations to ensure stability under varying environmental and sensory conditions. Additionally, it seeks to evaluate the potential of multi-task learning frameworks to utilize auxiliary tasks for more effective adaptation, ultimately enabling AI systems to generalize and perform reliably in diverse and challenging real-world applications.

Applicants to research degree programmes should normally have at least a first class or an upper second class British Bachelors Honours degree (or equivalent) in an appropriate discipline. The criteria for entry for some research degrees may be higher, for example, several faculties, also require a Masters degree. Applicants are advised to check with the relevant School prior to making an application. Applicants who are uncertain about the requirements for a particular research degree are advised to contact the School or Graduate School prior to making an application.

The minimum English language entry requirement for research postgraduate research study in the School of Computer Science is an IELTS of 6.5 overall with at least 6.5 in writing and at least 6.0 in reading, listening and speaking or equivalent. The test must be dated within two years of the start date of the course in order to be valid.