My name is Niall Twomey, and I'm a postdoctoral researcher on machine learning and data mining.
I will be updating this site regularly with posts/tutorials about my research, machine learning, and my publications. Additionally, I have added some projects that I have worked on under the "projects" tab on the navigation bar. These generally are either of personal interest to me, or they are associated to some software I have published on GitHub.
Since 2013, I have been a postdoctoral researcher in the SPHERE project at the University of Bristol, in the Department of Electrical Engineering, where I am also associated with the Intelligent Systems Lab. We work very closely with the Universities of Reading and Southampton, and we also have close ties to industry, including IBM and Toshiba. My research in the SPHERE project is on human-centred activity recognition for behavioural pattern detection for the purposes of healthcare assessment, and my day-to-day research is on structured prediction, transfer learning, active learning, and classifier calibration. Quantification of uncertainty underpins most of this research, and I find the Bayesian framework particularly elegant for this, and I am particularly emphatic about the need for sensible visualisation for theoretical and practical understanding of machine learning.
In 2012, I was a postdoc in the Embedded Systems Laboratory in the Department of Electronic Engineering at University College Cork where I was involved in designing, building and implementing body-area-network devices, digital signal processing algorithms, and machine learning in online, autonomous systems. My PhD research (2008-2012) was funded by an Enterprise Partnership Scheme between Intel and the Irish Research Council for Science and Engineering Technology (IRCSET), and I collaborated quite extensively with the Department of Paediatrics and Child Health for data collection. The objective of my research was to design and test the applicability of machine learning and artificial intelligence in the detection and prediction of pathological signatures of food allergy from cardiac physiology.
For those interested, the reason for selecting "otuama" for my web domain was due to the seeming absence of "twomey"-based domains names. Fortunately, "Ó Tuama" is the word for "Twomey" in the Irish language, and there were still some of these available.