Imperial College London Featured PhD Programmes
University of Southampton Featured PhD Programmes
University of Reading Featured PhD Programmes

Self-Supervised Learning and Variational Inference for adaptation in Deep Learning

  • Full or part time
  • Application Deadline
    Monday, June 01, 2020
  • Funded PhD Project (European/UK Students Only)
    Funded PhD Project (European/UK Students Only)

Project Description

We are looking for a highly motivated prospective PhD student to undergo a 3-year fully funded PhD project on Machine/Deep Learning foundations/theory and applications within the Department of Computing Science at the University of Aberdeen (UoA).

UoA is one of the oldest Universities in the UK with a strong focus on Artificial Intelligence, and consistently ranks among the top 200 Universities in the world.

Deep Learning theory has seen an immense development in the past few years across a number of areas, such as convolutional neural networks, capsule networks [1], generative adversarial networks, Bayesian deep learning [2], etc. However, some open problems such as improving predictive performance whilst reducing complexity, learning with few examples, explaining decisions, estimating uncertainty, improving routing and scaling in capsule networks are some areas than further investigations are needed to pass onto the next phase of deep/machine learning research.

This project aims at developing novel techniques in the areas of self-supervised learning, variational inference and domain adaptation, especially when it comes to dealing with few examples in real-world datasets that might be noisy, along with incorporating uncertainty for more explainable and effective decision-making process.

The idea of self-supervised learning is that one can aim at exploiting knowledge within the available data beyond the task at hand, i.e. enhancing the representation learning process via introducing pretext tasks, which can enhance the learning process and potentially the adaptation to other domains. In addition, many real-world applications require some sort of transparency in the decision-making process, therefore incorporating variational inference properties can be important to gaining trust on the deep learning outputs.

Stemming from the PI’s collaborations and current research activity, this PhD project will align and apply new theory in impactful application areas that include but are not limited to environmental datasets and energy [3], Oil and Gas industry (e.g. gas turbines) [4], nuclear reactors [3] and food industry [5].

There is some scope to shape the exact theoretical focus of this PhD project and title to align with the interests and background of the prospective student as well as the various collaborators that could be included in the project, such as the National Decommissioning Centre, Centre for Ecology and Hydrology and University of Lincoln.

You will work closely with other researchers and PhD students within Dr Leontidis’s lab on Applied Machine Learning, as well as other areas of the University that are considered of high strategic importance.

Selection will be made on the basis of academic merit. The successful candidate should have, or expect to obtain, a UK Honours degree at 2.1 or above (or equivalent) in Computer Science, Mathematics or Engineering.

Highly desirable - MSc (or near completion) in Machine Learning, Artificial Intelligence or Computer Science

Knowledge of: Machine Learning, Deep Neural Networks, Mathematics, Programming in Python.
Knowledge of Tensorflow, Keras, Pytorch and/or other deep learning frameworks would be advantageous

APPLICATION PROCEDURE:

• Apply for the Degree of Doctor of Philosophy in Computing Science
• State the name of the lead supervisor as the Name of Proposed Supervisor
• State the exact project title on the application form

Please include the following documentation when you apply:

*BSc and MSc Degree Certificates and Academic Transcripts
*2 Academic References
*Detailed CV
*Detailed personal statement indicating why you are interested in the project


Closing date for applications is 12 noon on 1st of June 2020, but we reserve the right to close the advert earlier should a suitable candidate be found.

Starting date: 1st of October 2020 or as soon as possible thereafter

Funding Notes

Tuition Fee waiver only, provided at UK/EU rates and stipend paid monthly in arrears (for 2019/2020 = £15,009). International students are welcome to apply, providing they can meet the difference between UK/EU and International tuition fees (2019/2020 = £15,680 per annum) from their own resources for the duration of study.

References

1. De Sousa Ribeiro, Fabio, Georgios Leontidis, and Stefanos Kollias. "Capsule Routing via Variational Bayes." AAAI, 2020.
2. Ribeiro, Fabio De Sousa, Francesco Calivá, Mark Swainson, Kjartan Gudmundsson, Georgios Leontidis, and Stefanos Kollias. "Deep Bayesian Self-Training." Neural Computing and Applications (2019): 1-17
3. Calivá, Francesco, Fabio Sousa De Ribeiro, Antonios Mylonakis, Christophe Demazi’ere, Paolo Vinai, Georgios Leontidis, and Stefanos Kollias. "A deep learning approach to anomaly detection in nuclear reactors." In 2018 International joint conference on neural networks (IJCNN), pp. 1-8. IEEE, 2018.
4. McGinty, Jason, Stefanos Kollias, and Georgios Leontidis. "Optimising remedial outcomes for Gas Turbines through large scale data analysis." in Industrial Maintenance and Reliability Manchester, UK 12-15 June, 2018 (2018): 85.
5. Onoufriou, George, Ronald Bickerton, Simon Pearson, and Georgios Leontidis. "Nemesyst: A hybrid parallelism deep learning-based framework applied for internet of things enabled food retailing refrigeration systems." Computers in Industry 113 (2019): 103133.

Email Now

Insert previous message below for editing? 
You haven’t included a message. Providing a specific message means universities will take your enquiry more seriously and helps them provide the information you need.
Why not add a message here
* required field
Send a copy to me for my own records.

Your enquiry has been emailed successfully





FindAPhD. Copyright 2005-2020
All rights reserved.