PTNO: Training Neural Operators with Noisy Monte Carlo Estimates for Particle Transport Problems

License: See LICENSE file · Updated 2026-09-30
My rating

PTNO is a neural operator designed for particle transport under multiple scattering, a problem relevant to radiative transfer and plasma physics. It learns surrogate models directly from noisy, low-cost Monte Carlo labels rather than requiring expensive, well-converged Monte Carlo solutions. The paper examines the challenges posed by this training-data regime.

Tags: