Applications are invited for a fully funded 3-year PhD studentship (tuition fee waiver + monthly stipend) in the School of Computer Science, Faculty of Science, University of Nottingham Malaysia Campus, under the supervision of Associate Professor Dr Iman Yi Liao.
A Novel Data Space Regularization Method for Solving Anti-Learning Problems in Machine Learning
In Machine Learning (ML), sample data of a specific problem are often used to train the machine to learn the behaviour of the data and predict the behaviour of any new data from the same problem. It generally works well when a large amount of sample data is available. However, the prediction accuracy drops severely when the sample size is small as compared to the dimension of the data and it often performs worse than random guessing. This is known as anti-learning, a common problem faced in contemporary research such as genomics, chemometrics, and image processing to name a few. Although in the current literature efforts have been made to adjust the bias caused by the small sample size data and different classifiers/predictors are combined to give better prediction accuracy, it does not change the fact that the data information is still limited by the sample data.
In this research we propose a novel method to tackle the problem with the main objective of increasing the data information by regularizing the data space. We take the analogy of curve interpolation, where a smooth curve can be reconstructed from a few sample points using a hidden structure although it is not observable from the sample points. Similarly yet less intuitively, we can represent any high dimensional data in a parametric/non-parametric way, build its hidden structure, with which the data space can be regularized from the sample data. The key point is that the hidden structure shall compensate for the information that any classifier/predictor can produce. New data can be mapped onto the regularized data space for classification/prediction using standard/advanced ML tools. We will test and validate the proposed data space regularization method with the applications to Forensic Cranio-facial Identification and colorectal cancer patients’ survival rate prediction.
The project is funded by the Fundamental Research Grant Scheme (FRGS), Ministry of Higher Education (MOHE), Malaysia.
According to FRGS requirement, Malaysian citizens with an MSc/MEng degree or a 1st class BSc/BEng degree, in Computer Science, Mathematics, Electrical and Electronic Engineering, or any other related area are welcome to apply. Previous experience in Machine Learning, Pattern Recognition and Computer Vision will be an advantage but not a must. The successful candidate is expected to take up the studentship in September 2014, or no later than December 2014.
For further enquiry on the project and the application procedure, please send your CV together with a cover letter to explain your career aim and why you want to apply, to:
Associate Professor Dr Iman Yi Liao
School of Computer Science,
Faculty of Science, The University of Nottingham Malaysia Campus
Selangor Darul Ehsan
Tel: +60-3-8725 3438
Posted on 11th August 2014