Emily O'Riordan
Research student,
- oriordane1@cardiff.ac.uk
- Room 2.02, 21-23 Senghennydd Road, Cathays, Cardiff, CF24 4AG
Overview
I am a final year PhD student in the School of Mathematics. My research spans several disciplines, but is largely focused in the cross-section between statistics and data science. I use methods from linear algebra and optimization to explore data analysis methods, particularly when applied to high-dimensional datasets.
My PhD has focused on the use of distance measures in high-dimensional spaces, and the development of new distance measures to counteract the so-called curse of dimensionality. The application of such research is broad, as distance measures are heavily relied upon in statistical, data science and machine learning methods.
Research
Research interests
- Multivariate data analysis
- Statistics
- Machine learning
- Optimization
Research Group
Teaching
I have taught on the following undergraduate courses:
- Computing for Mathematics
- Multivariate Data Analysis
I have also delivered material for the following masters courses:
- Statistics of Big Data
- Statistical Methods
- Summer School on Big Data and High Performance Computing
I am also the lead statistics tutor for the Maths Support service.
Thesis
Anomaly detection for large complex data
My PhD has focused on the use of distance measures in high-dimensional spaces, and the development of new distance measures to counteract the so-called curse of dimensionality. In high-dimensional settings, measuring dispersion of data is fraught with challenges. The novel distance measures I have produced during my PhD aim to consider the following issues:
- The (approximate) degeneracy of data, making the inverse of the covariance matrix non-existent
- The high-probability of correlation within groups of data, which can return unwanted results when measuring distances
- The possibility of different subspaces within one set of data
- The computational time of high-dimensional data methods, which often makes analysis very expensive, impractical or completely infeasible.
The application of such research is broad, as distance measures are heavily relied upon in statistical, data science and machine learning methods.
Funding source
CASE EPSRC
