Thomas Warford: A Random Walk to HMC

Hi everyone! My name is Thomas and this summer I’ll be working on scaling Hamiltonian Monte Carlo (HMC) with the Hartree Centre. In this post I’ll take you through the random walk that lead me to working on this project.

First, a little about me. I’m a Physics undergraduate at the University of Manchester with a strong interest in the application and physical limits of computation. In my spare time, I enjoy rock climbing (although I’ve yet to climb outdoors!) and playing rugby. This summer will be my first experience collaborating on code, and I’m excited to write code that other people will build upon.

Below is some of the past work I’ve done with monte carlo methods – namely simulating the propagation of neutrons through different materials. I also have experience with Linux due to my efforts to maximize cryptocurrency mining profit, which was rewarding in the sense it taught me a lot about computers. As a result of this experience, and my Physics background, I feel like this project is a great fit and I’m excited to see how far we can take it.

The hamiltonian monte carlo method is commonly used for estimating the parameters of a distribution, as well as for generating samples from a known distribution. Fascinatingly, as the name suggests, the method takes inspiration from Physics, with ‘energy’ and ‘momentum’ playing a key role in the algorithm. It’s strange that the analogy of a hockey puck sliding over a smooth hill is so applicable to statistics.

That’s all for now – come back in 2 weeks for a better overview of hamiltonian monte carlo – and some details of the parallel implementation!