About this project
This PhD project will investigate hierarchical decomposition of large neural networks into networks of subtasks to simulate the modularity found in human cognition problems in big data. Application areas are particularly in natural language processing. Neural networks can successfully learn complex behaviours at different layers of abstraction from small-scale connections between individual data points to large-scale connections between entire clusters of data points. Some input features will be more relevant to the model’s output predictions than others. The aim of this project is to identify these relevant features automatically during learning and partition large networks into networks of subtasks. This is crucial to the analysis of big data, where nature-inspired algorithmic selection of subtasks will lead to scalability, faster learning, and a better understanding of the natural decomposition of complex problems into sub-problems.
Lead Supervisor: Dr Nina Dethlefs, firstname.lastname@example.org
Prof Ken Hawick
Dr Alexander Turner
Full-time UK/EU PhD Scholarships will include fees at the ‘home/EU' student rate and maintenance (£14,121 in 2016/17) for three years, depending on satisfactory progress.
Full-time International Fee PhD Studentships will include full fees at the International student rate for three years, dependent on satisfactory progress.
Candidates should have excellent programming skills and a degree in Computer Science or a related discipline. Experience in machine learning is essential. Further knowledge and expertise of one or more of the following areas is highly desirable: evolutionary algorithms, deep learning, natural language processing, embedded systems and the Internet of Things (IoT).
Successful applicants will be informed of the award as soon as possible and by 8th May 2017 at the latest.
Find out more about research in Computer Science.