Reducing Frequency Bias of Fourier Neural Operators in 3D Seismic Wavefield Simulations Through Multistage Training

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Implementation of multistage training reduces frequency bias in seismic simulations, improving accuracy and reliability.

Multistage training reduces frequency bias in seismic simulations, enhancing results and accuracy.
The Fourier neural operator reduced errors on higher frequencies, improving overall ground-motion performance.
Analysis of the 3D Helmholtz wave equation showed improved learning at various frequencies with multistage training.
This approach indicates potential for broader applications in seismological modeling and simulations.

Kong et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark