Postdoctoral researcher, Johns Hopkins University, Department of Applied Mathematics and Statistics
I completed my double bachelors in mathematics and physics at the University of Wollongong (UoW) in 2014. I was awarded First Class Honours and a PhD in pure mathematics by UoW in 2015 and 2019 respectively under the supervision of Adam Rennie. For my Honours thesis, on the equivariant Atiyah-Singer index theorem, I was awarded the University Medal. For my PhD thesis, on the noncommutative geometry of foliated manifolds, I was awarded the Best Thesis Award in the Faculty of Engineering and Information Sciences.
I subsequently undertook a 4-month postdoc at the Australian National University and a 12-month postdoc at the University of Adelaide, both in pure mathematics, under the mentorship of Alan Carey and Mathai Varghese respectively. During this time, I continued my research into the topology and geometry of foliated manifolds, culminating in the systematisation of the construction of holonomy groupoids for foliations and the construction of a Chern-Weil theory for singular foliations.
In May 2021, I joined the Australian Institute for Machine Learning as a member of Simon Lucey's research group. Over the following two years I developed a research program tackling theoretical aspects of deep learning including equivariance, optimization, and generalization. My work has been published at conferences such as NeurIPS, ICML, and CVPR.
As of July 2023, I am continuing my research as a joint member of Rene Vidal's deep learning theory research group, first at Johns Hopkins University and since July 2024 at the University of Pennsylvania. My current research centers on the construction of general theoretical foundations for deep learning which can explain the miraculous empirical performance of deep learning models.
I spend my spare time exploring Baltimore with my partner and reading philosophy, especially the work of Kant, Fichte, Schelling, and Hegel.
Email: lemacdonald@protonmail.com