In understanding natural systems over hundreds of years, physicists have developed a wealth of dynamics and viewpoints. Some of these methods, when viewed (and abstracted) through the lens of computation, could lead to new optimization and sampling techniques for problems in machine learning, statistics, and theoretical computer science. I will present a couple of recent examples from my research on such interactions between physics and algorithms – a Hamiltonian dynamics inspired algorithm for sampling from continuous distributions and a Boltzmann’s equation based algorithm for estimating the partition function for discrete distributions.

Based on joint works with Oren Mangoubi and Damian Straszak.