Ethan C. Jackson, PhD

Post-doctoral research fellow at the University of Guelph Machine Learning Research Group and Vector Institute interested in algebraic methods in AI, quality diversity algorithms, and bias mitigation.

Education and Academic Affiliations

  • Post-Doctoral Fellow, University of Guelph and Vector Institute (2019 – Present)
  • Postgraduate Affiliate, Vector Institute (2018 – Present)
  • PhD Computer Science, University of Western Ontario (2014 – 2019)
  • MSc Computer Science, Brock University (2012 – 2014)
  • Honours BSc Computer Science and Mathematics, Brock University (2008 – 2012)

Research Highlights

Algebraic Structures for Neural Computation
Is the structure of a neural network more than just its graph? In current research, I am further investigating the interpretation of neural computation using algebraic structures, including relation and abstract matrix algebras. Important applications include differentiable architecture mutability and neural plasticity.

Decision-Directed Data Decomposition
Bias is pervasive in machine learning. In this work, we introduce a fast algorithm for debiasing data, including neural representations or embeddings, using iterative projections onto decision boundaries.

Action-Based Novelty Search in Evolutionary Deep Reinforcement Learning
How do we want artificial agents to behave? Novelty search is a powerful framework for reinforcement learning demonstrating that optimal performance is not always achieved by optimizing for reward. In this work, we introduce a general formulation of novelty search that promotes behavioural diversity by comparing entire sequences of actions.

Mutable Network Architectures in Evolutionary Deep Reinforcement Learning
Convolutional neural networks are highly effective neural network modules for modelling sensory perception. What neural modules are best suited for modelling complex behaviours? In this work, we contribute to the reemergence of evolutionary algorithms as powerful tools for reinforcement learning by combining deep convolutional neural networks with architecturally mutable subnetworks.

Full List of Publications


Teaching

I am enthusiastic about fostering a highly interactive classroom. At the University of Western Ontario, I have delivered 4 courses as an instructor, worked with Jody Culham at the Brain and Mind Institute to develop graduate-level interactive tutorials on neuroimaging data analysis, and have guest lectured on artificial neural networks.

Sessional Instructor at the University of Western Ontario

  • COMPSCI 2120A – Coding Essentials (Fall 2018)
  • COMPSCI 1026A – Computer Science Fundamentals I (Summer 2017)
  • COMPSCI 2212B – Software Engineering (Winter 2017)
  • COMPSCI 1027A – Computer Science Fundamentals II (Fall 2016)

Course Development Experience

  • PSYCHOLOGY 9223 – Neuroimaging of Cognition (2017 – 2018)

Guest Lectures

  • Neural Nets Week – COMPSCI 4414A – Intro to Data Science (2019)
  • Deep Dive on Artificial Neural Networks – COMPSCI 4414A – Intro to Data Science (2018)

Contact

I can be contacted by email at jackson <dot> ethan <dot> c <at> gmail <dot> com or via any of the following social links.