About this project
This project is in the area of deep learning and investigates memory models in the context of big data. Memory models (or MemNNs) are specific types of neural networks that compute output predictions not just based on a compressed hidden state (as regular neural nets) but takes a “memorised” representation of the input into account as well. This gives them two advantages: (1) they are able to process and learn with longer input sequences than conventional (recurrent) neural networks, and (2) they can give us some insight into the features that led to a certain output prediction and thus provide rationales for their predictions.
This project can take two directions. It can focus on explainable deep learning and find ways to highlight the input features that gave rise to the output predictions. This will provide a better understanding of the network’s “reasoning”. The project can alternatively focus on efficient processing of big data to enable application of MemNNs - which tend to be computationally expensive - in e.g. real-time settings, using the University’s HPC Viper. While this is the general gist of the project, there is room for adaptation to individual research interests of the successful candidate.
Dr Nina Dethlefs - email@example.com
Full-time UK/EU MSc Scholarship will include fees at the ‘home/EU' student rate and maintenance (£14,553 in 2018/19) for one year until the end of August 2019.
Applicants should have at least a 2.1 undergraduate degree in Computer Science or related discipline. Excellent programming skills and an interest in machine learning and AI are essential criteria. Experience in the area of machine learning and data mining is an advantage.
How to apply
Applications for scholarship consideration at the University of Hull should be made through the Postgraduate Application system.
On the second page of your application, please select “Graduate Scholarship” as the type of scholarship you are applying for.
Applicants are strongly encouraged to first identify and contact a potential supervisor.
Application deadline: Monday 26 March 2018