Particle transport neural operator speeds simulation with noisy training
PTNO: Training Neural Operators with Noisy Monte Carlo Estimates for Particle Transport Problems
Artificial IntelligenceMachine Learning
Summary
Simulating how particles move and scatter is important but very slow when relying on traditional methods that trace many particles. The authors propose a neural network model called PTNO that learns from quick but noisy simulations instead of precise, costly ones. Their method handles noisy data better by using a special loss function and output design, allowing fast and accurate predictions. This makes simulating particle behavior in things like fusion reactors or light passing through materials much faster and cheaper than older methods.
What this means in practice
- •For fusion reactor engineers: Speed up neutron transport simulations needed for reactor design and safety analysis significantly.
- •For computer graphics developers: Accelerate radiative transfer calculations in rendering participating media for realistic light effects.
Authors
Yubo Cao, Xi Deng, Mengqi Xia, Vignesh Gopakumar, Ander Gray, Anima Anandkumar
Abstract
Particle transport under multiple scattering is central to radiative transfer and plasma physics, yet high-fidelity Monte Carlo (MC) simulations must trace prohibitively many particles. Learning-based surrogates can amortize this cost, but typically train on expensive, well-converged MC solutions. We propose the Particle Transport Neural Operator (PTNO), a neural operator that learns particle transport surrogates directly from noisy, low-cost MC labels. Such labels pose two challenges: (1) high variance, which destabilizes standard supervised learning, and (2) a high dynamic range (HDR) spanning many orders of magnitude. For the first, we learn the solution operator from noisy labels of many configurations, amortizing MC cost and generalizing to unseen configurations. Because MC labels are unbiased, we show that the squared loss on them shares its minimizer with the loss on converged solutions, and our budget-allocation study over training scenes $M$, MC samples per render $N$, and independent renders per scene $K$ shows that many noisy scenes beat fewer converged ones. For the second, a nonlinear transform such as the logarithm biases noisy supervision. Instead, PTNO keeps labels in physical space and enforces positivity with a softplus output layer that represents small values effectively. We further train with a pointwise relative $L_2$ loss (PRelL2), the stop-gradient relative loss of HDR denoising and neural rendering, which normalizes each residual by the stop-gradient prediction instead of the noisy label. We demonstrate PTNO on neutron transport in fusion reactors and radiative transfer in participating media. On the two neutronics tasks, PTNO is $10^4$-$10^5\times$ faster than converged MC on the same CPU and $10^3$-$10^5\times$ cheaper than MC at matched accuracy; on the two radiative-transfer tasks, MC at matched accuracy costs $0.8$-$11\times$ as much as PTNO.