Transfer Learning of High-Fidelity Opacity Spectra in Autoencoders and Surrogate Models

TL;DR

This paper demonstrates that transfer learning enables neural networks to accurately reproduce high-fidelity opacity spectra in high energy density physics simulations using significantly fewer high-fidelity samples, reducing computational costs.

Contribution

It introduces a transfer learning approach that allows neural networks to generate high-fidelity spectra with minimal high-quality training data, improving efficiency in physics simulations.

Findings

Median errors of 3-4% in spectra reproduction

Achieved high accuracy with only 50 high-fidelity samples

Transfer learning from low- to high-fidelity data is effective

Abstract

Simulations of high energy density physics are expensive, largely in part for the need to produce non-local thermodynamic equilibrium opacities. High-fidelity spectra may reveal new physics in the simulations not seen with low-fidelity spectra, but the cost of these simulations also scale with the level of fidelity of the opacities being used. Neural networks are capable of reproducing these spectra, but neural networks need data to to train them which limits the level of fidelity of the training data. This paper demonstrates that it is possible to reproduce high-fidelity spectra with median errors in the realm of 3\% to 4\% using as few as 50 samples of high-fidelity Krypton data by performing transfer learning on a neural network trained on many times more low-fidelity data.