[This website is from my time as a PhD student. It is no longer actively maintained.]

Hi, I'm Tom! I started my DPhil in Computer Science at Oxford University in 2019, and I'm part of the Future of Humanity Insitute DPhil Scholars program at Oxford.

My thesis is concerned with understanding how one can build a system, using well-understood theory and algorithmic reductions, together with more applied (but less well understood) practical Machine Learning techniques, for building practical and robust lifelong learning agents.

Email: thomas (dot) orton (at) cs (dot) ox (dot) ac (dot) uk

Links to some projects and notes:

  1. Computationally Bounded Rationality From Three Perspectives: Precomputation, Regret Tradeoffs, And Lifelong Learning (PhD Thesis)
  2. Trading Off Resource Budgets For Improved Regret Bounds (NeurIPS '22)
  3. Modeling Precomputation In Games Played Under Computational Constraints (IJCAI '21)
  4. Senior Thesis: Payment Does Not Imply Consensus
  5. A blog post on Belief Propagation for the Stochastic Block Model
  6. Capturing Entropy of Locally Dependent Functions
  7. Talk notes on the paper "Stronger generalization bounds for deep nets via a compression approach" (S. Arora et al.)
  8. A survey on regret minimization for financial portfolios
  9. A GAN model for upscaling Anime Videos
    10. Before and after conversion.

Other:

  1. Some of my orchestral compositions
  2. I competed in programming olympiads in high school, and occasionally set problems for South African national training camps/competition rounds. My favourite problem I set is towers (problem C)