Academic Work

My prior research revolves primarily around the generalization of artificial neural networks. My most recent work looked at the ability of linear neural networks to systematically generalize. We investigated the impact of dataset structure, architecture and learning regime on this ability and found conditions when systematic generalization could occur.

For my Masters work I looked to create selective plasticity regularizers which were based on the geometry of the loss landscape. By using the metric tensor of this landscape we were able to create regularizers which would adapt their strength based on the individual importance of a parameter to the network as a whole.

Education

• Bachelor of Science (with Distinction) in Computer Science and Economic Science, 2016 - 2018
• Bachelor of Science Honours (with Distinction) in Computer Science, 2019
• Master of Science (with Distinction) in Computer Science, 2020-2021

Academic Writing

• Masters Dissertation

Academic Positions

• Head Academic Tutor at the University of the Witwatersrand, 2020-2021. Courses:
  • Machine Learning III
  • Computer Vision IV
  • Natural Language Processing IV
  • Mathematical Foundations of Data Science IV
• Academic Tutor at the University of the Witwatersrand, 2019. Courses:
  • Introduction to Databases II
  • Analysis of Algorithms II