Matt L. Sampson

Matt Sampson
Physics for AI
AI for Physics
PhD candidate at Princeton
Hey, I’m Matt, a PhD researcher at Princeton University working with Prof. Peter Melchior. I research ways to advance machine learning and science, particularly focusing on optimization, representation learning, and dynamical systems. Physics allows us to uncover order from chaos. My aim is to bring this approach to the field of machine learning optimization because no matter how complex and advanced our models become, they are only as useful as our ability to train them.
Previously I was at the Australian National University where I performed numerical experiments and helped develop code to simulate cosmic ray propagation through a dynamically evolving turbulent plasma.
Long-term vision: To design optimization methods that make it possible to build and train models with a deep understanding of complex dynamical systems — enabling new scientific discoveries and forming the foundations for increasingly general, physics-inspired intelligence.
news
May 13, 2025 | Lucky to be one of 4 students from Princeton University nominated for the 2025 Google PhD Fellowship! The first nomination from the astrophysical sciences deptartment at Princeton |
---|
selected publications
- Path-minimizing latent ODEs for improved extrapolation and inferenceMachine Learning: Science and Technology, Jun 2025
- Cosmic ray and plasma coupling for isothermal supersonic turbulence in the magnetized interstellar mediumarXiv e-prints, Jun 2025
- Score-matching neural networks for improved multi-band source separationAstronomy and Computing, Oct 2024
- Spotting Hallucinations in Inverse Problems with Data-Driven PriorsIn ICML ML4Astrophysics Workshop (Oral), Jul 2023