PINNup: Robust neural network wavefield solutions using frequency upscaling and neuron splitting

TL;DR

This paper introduces PINNup, a novel physics-informed neural network approach that employs frequency upscaling and neuron splitting to efficiently and accurately model high-frequency wavefields in seismic applications.

Contribution

The paper presents a new PINN implementation that grows in complexity with frequency, leveraging pre-trained models to improve high-frequency wavefield solutions.

Findings

PINNup outperforms traditional PINN in convergence speed and accuracy.

It achieves high-frequency wavefield solutions with a simple two-hidden-layer network.

The method reduces training costs for high-frequency seismic modeling.

Abstract

Solving for the frequency-domain scattered wavefield via physics-informed neural network (PINN) has great potential in seismic modeling and inversion. However, when dealing with high-frequency wavefields, its accuracy and training cost limits its applications. Thus, we propose a novel implementation of PINN using frequency upscaling and neuron splitting, which allows the neural network model to grow in size as we increase the frequency while leveraging the information from the pre-trained model for lower-frequency wavefields, resulting in fast convergence to high-accuracy solutions. Numerical results show that, compared to the commonly used PINN with random initialization, the proposed PINN exhibits notable superiority in terms of convergence and accuracy and can achieve neuron based high-frequency wavefield solutions with a two-hidden-layer model.