Real-world environments in today’s highly interconnected digital world are often non-stationary. Learning systems such as classifiers used for pattern recognition, make the stationarity assumption in input data distribution density during training and operation phases. Their field performance may become unsatisfactory and therefore such learning systems may have limited practical use. A very good example of a non-stationary environment is an magnetoencephalography and electroencephalography (M/EEG) based brain-computer interface (BCI) in which a communication link is established between a computer and the user through a process involving thinking/imagining a set of cognitive tasks.
In this, a classifier is used to detect the electrophysiological correlates, resulting from the cognitive tasks performed by the user [1-4]. The non-stationary characteristics of M/EEG has been one of the primary impediments in developing a practically useful BCI, despite decades of R&D work. In fact prevalence of such non-stationary systems is wide-spread, e.g., autonomous robots, share price trend predictions, spam filtering, surveillance, and bioinformatics. In all these areas, a learning system devised under the assumption of input data distribution invariance results in suboptimal performance while the system is being used either as a classifier or a predictive model.
There is thus an urgent need of developing reliable techniques for monitoring, predicting, categorising and often controlling non-stationary systems’ characteristics in real-world systems.
To account for EEG non-stationarities appearing as covariate shifts, we have recently developed a covariate shift detection and adaptation methodology using an Exponentially Weighted Moving Average (EWMA) modelling strategy and successfully applied in a motor imagery based BCI [1-4]. Advancing this work further, this project aims to investigate how characteristics of non-stationary systems be effectively learned on-line in the absence of manual labelling of the streaming data.
It will involve investigating a range of approaches applied in multiple applications in a collaborative way. It may involve developing techniques for learning from the changes in inputs alone, possibly transductively based on the previous data; devising adaptation strategies that can make effective use of the available information to extract knowledge about the system variability and also making continuous update to account for the new information without forgetting already learnt information possibly using prediction error minimisation (PEM) based active inferencing and Bayesian belief propagation approach.
If the University receives a large number of applicants for the project, the following desirable criteria may be applied to shortlist applicants for interview.
Vice Chancellors Research Scholarships (VCRS)
The scholarships will cover tuition fees and a maintenance award of £14,777 per annum for three years (subject to satisfactory academic performance). Applications are invited from UK, European Union and overseas students.
The scholarship will cover tuition fees at the Home rate and a maintenance allowance of £ 14,777 per annum for three years. EU applicants will only be eligible for the fees component of the studentship (no maintenance award is provided). For Non EU nationals the candidate must be "settled" in the UK.
As Senior Engineering Manager of Analytics at Seagate Technology I utilise the learning from my PhD ever day
Adrian Johnston - PhD in InformaticsWatch Video
Monday 18 February 2019
19-20 March 2019
A key player in the economy of the north west
When applying for this PhD opportunity please quote reference number: