Deep Learning for Pattern Recognition Identification and Classification from Big Datasets

Applications are invited for a three year PhD to commence in October 2020 or February 2021.

The PhD will be based in the Faculty of Technology, and will be supervised by Dr. Ivan Jordanov and Prof. Raymond Lee.

Investigating and applying Deep Learning approaches for solving image, object, and signal recognition and classification problems. Analysis and design of efficient convolutional neural network (CNN) and long-short term memory neural network (LSTM) topologies and architectures and proposing learning/training, testing, and evaluation strategies for the adopted deep learning approaches and the employed neural network architectures on real world datasets.

In the recent years, Deep Learning (DL) has demonstrated remarkable success in solving image, object, and especially in speech and signal recognition systems. The current advancements of the deep learning approaches (DLA) provided evidence that on big data, sophisticated algorithms can achieve better performance than simple models (the traditional shallow learning methods). The DL ability to learn feature hierarchies with multiple levels of abstraction allows the system to learn complex functions while mapping the input to the output directly from data, without the need and complete dependency on man-crafted feature and high level concept extraction.

In your research you will aim at getting better and in-depth understanding of the working mechanisms behind the success of Deep Learning and will have to investigate DL approaches and algorithms, and propose suitable topologies and architectures for convolutional neural networks (CNN) when solving pattern recognition and classification problems and long-short term memory neural networks (LSTM) when dealing with time series and signal processing problems. You will also have to investigate and work on associated preprocessing data analytics problems, typical for most real world datasets and related to the data quality (e.g., how to deal with missing, incomplete, imbalanced, noised, shifted, etc. data).

The investigation will include designing learning/training, testing, and evaluation strategies for your deep learning approaches and the employed neural network architectures on real world datasets. For example, the research will lead to applying the developed system for gaining insights from big datasets collected during fetal monitoring at labor, achieving objective assessment and reducing the risk of new-born babies’ asphyxiation and brain damage, when inferring and predicting the labor outcome (as part of ongoing academic research collaboration with Oxford Centre for Fetal Monitoring Technologies, Nuffield Department of Women's and Reproductive Health, and The Big Data Institute, University of Oxford)

Fees and funding

Visit the research subject area page for fees and funding information for this project.

Funding availability: Self-funded PhD students only. 

PhD full-time and part-time courses are eligible for the UK Government Doctoral Loan (UK and EU students only).