Title: Explainable AI for computational Algebra

Abstract: A Computer Algebra System is a piece of software that performs symbolic mathematical computations with exact precision, as a human mathematician would by hand. These tools are used regularly within science, industry and education. There has been increasing interest in the use of machine learning (ML) within such mathematical software since many of the algorithms within come with non-critical choices to be optimised. I.e., they have choices which, although having no effect on the mathematical correctness of the output, can greatly affect output presentation and the resources required to produce output.
Recently, there has been some evidence that Explainable AI techniques (i.e. ML models which are either designed to be interpretable by a human, or can be analysed subsequently to explain their behaviour), may be used to identify new mathematical knowledge. This project seeks to establish whether and how such tools could be used in a structured way in the development of computer algebra systems. This project focuses in particular on overcoming current barriers on synthetic mathematics data generation.

Supervisors: 

1. Dr. Mathew England (Coventry University)
2. Dr. Nayyar Zaidi (Deakin University)
3. Dr. James Brusey (Coventry University)
4. Dr. Dhananjay Thiruvady (Deakin University)
5. Prof. John Yearwood (Deakin University)

Uzma is a PhD student in computer science, specializing in explainable AI within computational algebra.  She is currently pursuing a dual (cotutelle) Ph.D. degree at the Centre for Computational Science and Mathematical Modelling of Coventry University, UK, and the School of Information Technology of Deakin University, Australia. 

Uzma received her bachelors degree in Electrical Engineering from National University of Sciences and TEchnology (NUST). She recieved her masters degree in Biomedical Engineering from NUST as well. Her research interests are in Machine Learning, Computer Algebra, and Explainable AI.

Previous Research Experience:

Human Systems Lab (SMME_NUST)

• Modelling a Subvocal Speech detection and recognition system using IMU and EMG.
• Pattern recognition for limb motion using time-frequency domain feature engineering.
• Denoising EMG signals using Variational mode decomposition.