This PhD Studentship is sponsored by Thales UK and EPSRC. The successful candidate will seek to solve the problems faced by applying sparse data learning techniques for detection and classification to domains with very limited labelled data. In Particular, the student will investigate Anomaly Detection, Few-Shot and Zero-Shot learning for object/ event classification, detection and segmentation, focussing on how these techniques can be applied to domains with little or no existing labelled data. The problem of anomaly detection has various applications which involve novel object or event detection in a large amount of input data. Conventional machine learning and deep learning algorithms have been examined for this task and some promising results have been reported, however, they often rely on access to enough training data, e.g. SVDD (Support Vector Data Description) and Deep-SVDD. Further research is needed into how to successfully conduct the task, if we have limited data, or little/no labelled data. This project will incorporate ideas from Few-Shot learning to allow the updating of classification, detection and segmentation models with new classes from very few examples or only a description in the case of Zero-Shot learning. By combining these ideas with anomaly detection the student will develop novel algorithms which allow the identification of the objects/events of interest at any one time, while ignoring uninteresting and normal data.
The PhD student will focus on the underlying algorithms applying the developed algorithms to popular open-source data-sets. The developed algorithms will not only aim to produce state-of-the-art results on these data but also investigate how they can be practically applied to assist in data analysis and decision making. In order to demonstrate this, the industrial partner of this project Thales UK, will provide a number of use-cases and data which will be used to demonstrate the scalability and accuracy of the algorithms in practice. This will include a broad range of use-cases in areas such as transportation, surveillance, and cyber-security.
The student will be sponsored by EPSRC and Thales UK and will work closely with the industry sponsor and have the opportunity to spend at least 3 months on placement with the business. . This will provide a fantastic route to applying the developed algorithms to real world products and generate a significant impact and the student will be expected to attend meetings with and present their work to the industrial sponsor.
The student will also collaborate with researchers and investigators in the University Defence Research Collaborations (UDRC), which is a multi-university EPSRC/Dstl project, aiming to develop underpinning signal processing and machine learning methods for defence and security future applications. As a result they will benefit from training programs, e.g. UDRC summer school, and knowledge exchange with various industrial partners and DSTL (Defence Science and Technology Lab).
The studentship comes with an annual stipend of £18,000 plus fees.
Applicants should have a first or second class UK honours degree or equivalent in a related discipline, such as electrical engineering, computer science, applied mathematics, statics or relevant degrees to the project. To be eligible for this funding, applicants must be a UK national or an EU national with settled status.
Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possibly supported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.
Funding (tuition fees and stipend) is available for UK students and EU students who have lived in the UK for 3+ years. EU students who have not lived in the UK for 3+ years are only eligible for tuition fees and not stipend.
International students are welcome to apply but external funding must be sourced – please provide details of the funding source you intend to apply to, or details of the funding you have already secured in your application.