I am super passionate about artificial intelligence and, more specifically, reinforcement learning. In particular, I am interested in designing agents capable of acting in a wide variety of environments. My current research focuses on autonomously learning hierarchies of abstractions that can be transferred between tasks. This parallels the way humans are able to transfer abstract concepts from one task to another in the form of symbolic representations.
Lectured COMS1018A: Introduction to Algorithms & Programming, consisting of ±350 students.
Guest lectured COMS4047A: Special Topics, covering reinforcement learning for an honours-level class of ±45 students.
Lectured COMS1018: Introduction to Algorithms & Programming, consisting of ±250 students.
Responsible for delivering Android applications related to the insurance industry, and integrating said applications into the existing backend and proprietary hardware systems.
Part of a small R&D team responsible for rewriting components of an ERP suite, including replacing the existing Telnet connection to the server with SSL, as well as providing a mechanism for session recovery should the connection be lost. Also responsible for rewriting the entire front-end in JavaFX to be used on multiple platforms.
Responsible for setting and conducting tutorials and lab sessions, as well as marking test scripts.
Learning Portable Representations for High-Level Planning
Proceedings of the Thirty-seventh International Conference on Machine Learning
If dropout limits trainable depth, does critical initialisation still matter? A large-scale statistical analysis on ReLU networks
Pattern Recognition Letters, Elsevier
Composing Value Functions in Reinforcement Learning
Proceedings of the Thirty-sixth International Conference on Machine Learning
An Analysis of Monte Carlo Tree Search
Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence
Local Publications (Refereed)
Quantisation and Pruning for Neural Network Compression and Regularisation
Learning Options from Demonstration using Skill Segmentation
Inter- and Intra-domain Knowledge Transfer for Related Tasks in Deep Character Recognition
Workshops, Symposia & Tech Reports
Learning Object-Centric Representations for High-Level Planning in Minecraft
Object-Oriented Learning (OOL): Perception, Representation, and Reasoning. Workshop at ICML.
Logical Composition for Lifelong Reinforcement Learning
4th Lifelong Learning Workshop at ICML 2020
A Boolean Task Algebra for Reinforcement Learning
Beyond "Tabula Rasa" in Reinforcement Learning (BeTR-RL): Agents that remember, adapt, and generalize. Workshop at ICLR.
Multi-Pass Q-Networks for Deep Reinforcement Learning with Parameterised Action Spaces
arXiv preprint arXiv:1905.04388
Learning to Plan with Portable Symbols
ICML/IJCAI/AAMAS 2018 Workshop on Planning and Learning
Will it Blend? Composing Value Functions in Reinforcement Learning
The 2nd Lifelong Learning: A Reinforcement Learning Approach (LLARLA) Workshop, FAIM 2018.
Learning Portable Symbolic Representations
2018 IJCAI Doctoral Consortium (extended abstract)
An Investigation into the Effectiveness of Heavy Rollouts in UCT
General Intelligence in Game-Playing Agents (GIGA’16) Workshop at IJCAI
The Effect of Simulation Bias on Action Selection in Monte Carlo Tree Search
University of the Witwatersrand (MSc thesis)
PhD in Computer ScienceUniversity of the Witwatersrand2017 - Present
MSc in Computer Science
(with distinction)University of the Witwatersrand2014 - 2016
BSc (Hons) in Computer Science
(with distinction)University of the Witwatersrand2013
BSc in Computer Science and Applied Mathematics
(with distinction)University of the Witwatersrand2010 - 2012
Google PhD Fellowship
Awarded the 2018 Google Africa PhD Fellowship in the field of Machine Learning.
PVT Educational Bursary
One of two recipients of the PV Tobias Educational Bursary for academically excellent candidates.
Chancellor's Gold MedalAwarded for the most distinguished graduate of the year.
- English (Native)
- Afrikaans (Passable)
Things I Like
- One true brace style
- Bitstream Charter
- Following the gradient
- Céline Dion
- The word "argillaceous"
Things I Dislike
- Incorrect gerund usage
- Not following the gradient
- Mispronunciation of "Euler"
Making meaningless, vague bars
- Artificial Intelligence
- Reinforcement Learning
- Transfer Learning
- Basic Arithmetic