Empowering Vision Transformers: A Self-Supervised Approach for Long-Tail Data Challenges

Summary

Self-Supervised Learning (SSL) is revolutionizing AI by reducing the dependence on labeled data, particularly benefiting applications in resource-constrained environments. This PhD research focuses on advancing Vision Transformers (ViTs) through SSL to tackle two pressing challenges: the imbalance of data distribution (long-tail) and the limitations it imposes on model generalization. By using SSL-based learning, this project aims to enhance ViTs’ generalisation and performance in underrepresented data categories, offering critical insights for more robust models in real-world applications.

Combining self-supervised learning (SSL) with Generative Adversarial Networks (GANs) effectively addresses class imbalance in datasets. SSL helps models learn robust representations from unlabeled data, which can enhance performance on minority classes. GANs generate synthetic samples to augment underrepresented classes, helping to balance the dataset. This combination allows for better generalization by fine-tuning classifiers on a richer, more balanced dataset. Ultimately, this integrated approach improves model performance, particularly in scenarios with significant class imbalance.

The proposal's primary objective is to enhance the generalisation and performance of ViTs through innovative SSL strategies in the context of long-tail data distribution. The framework will involve analysing long-tail data characteristics, exploring SSL methodologies such as contrastive learning and masked image modelling, developing GAN model for synthetic data generation, and evaluating the enhanced ViTs' performance on benchmark datasets focusing on metrics like accuracy and robustness. This research aims to provide insights into integrating SSL and GAN with ViTs, ultimately leading to robust AI models capable of doing the effective classification task.

In conclusion, this proposal seeks to innovate Vision Transformers for classification task through SSL and GAN to overcome challenges posed by long-tail data distributions, enhancing model generalisation while promoting fairness and sustainability in resource use.

The School of Computing at Ulster University holds Athena Swan Bronze Award since 2016 and is committed to promote and advance gender equality in Higher Education. We particularly welcome female applicants, as they are under-represented within the School.

Essential criteria

Applicants should hold, or expect to obtain, a First or Upper Second Class Honours Degree in a subject relevant to the proposed area of study.

We may also consider applications from those who hold equivalent qualifications, for example, a Lower Second Class Honours Degree plus a Master’s Degree with Distinction.

In exceptional circumstances, the University may consider a portfolio of evidence from applicants who have appropriate professional experience which is equivalent to the learning outcomes of an Honours degree in lieu of academic qualifications.

• Sound understanding of subject area as evidenced by a comprehensive research proposal
• A demonstrable interest in the research area associated with the studentship

Desirable Criteria

If the University receives a large number of applicants for the project, the following desirable criteria may be applied to shortlist applicants for interview.

• First Class Honours (1st) Degree
• Masters at 70%
• For VCRS Awards, Masters at 75%
• Experience using research methods or other approaches relevant to the subject domain
• Work experience relevant to the proposed project
• Publications - peer-reviewed
• Experience of presentation of research findings

Equal Opportunities

The University is an equal opportunities employer and welcomes applicants from all sections of the community, particularly from those with disabilities.

Appointment will be made on merit.

Funding and eligibility

This project is funded by:

• Department for the Economy (DfE)
• Vice Chancellor's Research Scholarship (VCRS)

Our fully funded PhD scholarships will cover tuition fees and provide a maintenance allowance of £19,237 (tbc) per annum for three years (subject to satisfactory academic performance).  A Research Training Support Grant (RTSG) of £900 per annum is also available.

These scholarships, funded via the Department for the Economy (DfE) and the Vice Chancellor’s Research Scholarships (VCRS), are open to applicants worldwide, regardless of residency or domicile.

Applicants who already hold a doctoral degree or who have been registered on a programme of research leading to the award of a doctoral degree on a full-time basis for more than one year (or part-time equivalent) are NOT eligible to apply for an award.

Due consideration should be given to financing your studies.