My PhD project focuses on explaining black-box classifications. More precisely on how logical rules can accurately represent the behaviour of a neural network and then be used to explain that behaviour.
I chose to do my PhD as part of this CDT because my project seemed to align with the objectives of the STAI CDT: combining safe and trusted (symbolic) learning techniques with data-driven (sub-symbolic) techniques, for a safer and more trustworthy approach to data-driven learning. Having a cohort makes the PhD far more social. The masterclasses offered by the CDT also have a very positive impact on one’s wider knowledge of the AI field. The Journal club is also a great opportunity for training one’s presentation skills. Listening to colleagues present at the Journal club helps to alert you of recent and important papers and concepts out of your specialised research area.
Research Experience: I have worked on solving tractable instances of integer programming (IP). It’s a clustering algorithm for solving the objective function in an IP model. The algorithm was based on unpublished preliminary work done by my supervisor – Dr Christopher Hampson. I have extended his work during my final year project to include negative weights and evaluation of clusters with internal bound > 1.
Work Experience: I have worked as a Software Engineer in Navico and Salesforce. I have also worked as a teaching assistant in Artificial Intelligence Planning, Programming in C++ and Scala (Practical Experiences of Programming), as well as Foundations of Computing 2.