Spencer Hill

Hi there! I’m Spencer Hill, a master’s student in the Department of Mathematics and Statistics at Queen’s University. I work on machine learning and information theory under the supervision of Fady Alajaji and Tamás Linder. I hold a BASc (Engineering) degree from Queen’s University, where I studied Mathematics and Engineering with an option in computing and communications (read: I was a mathematician with a coding side hustle).

I am broadly interested in problems that sit at the intersection of machine learning/artificial intelligence and mathematics. I am especially interested in investigating the theoretical limits of machine learning algorithms and designing ones which have provable guarantees (and can actually be implemented on a computer!). This has previously manifested as work on faster Guassian processes for computational chemistry, GANs, Quantum Machine Learning, neural image compression, and reinforcement learning.

Recently, my master’s research has focused on the theory of channel simulation, which is the problem of communicating a random sample from a distribution at low cost. This idea has recently been applied to neural-network-based compression to make the system end-to-end differentiable and able to be trained using gradient descent. My research (so far) has generalized the theoretical limits of channel simulation to an exponential cost.

I am also currently collaborating with researchers from the Queen’s University Smith School of Business and Natural Resources Canada to design reinforcement learning algorithms to optimally track the spread of invasive species.

For more information, check out my CV, publications, projects, or send me an email!