DiffPINN: Generative diffusion-initialized physics-informed neural networks for accelerating seismic wavefield representation

TL;DR

DiffPINN introduces a diffusion-based approach to rapidly initialize physics-informed neural networks for seismic wavefield modeling, significantly reducing training time while maintaining accuracy across diverse velocity models.

Contribution

The paper presents a novel latent diffusion strategy to efficiently initialize PINNs, enabling faster training for seismic wavefield simulations across various velocity models.

Findings

Accelerates PINN training by leveraging diffusion-generated parameter initialization.

Maintains high accuracy in both in-distribution and out-of-distribution velocity scenarios.

Demonstrates significant speedup compared to traditional PINN training methods.

Abstract

Physics-informed neural networks (PINNs) offer a powerful framework for seismic wavefield modeling, yet they typically require time-consuming retraining when applied to different velocity models. Moreover, their training can suffer from slow convergence due to the complexity of of the wavefield solution. To address these challenges, we introduce a latent diffusion-based strategy for rapid and effective PINN initialization. First, we train multiple PINNs to represent frequency-domain scattered wavefields for various velocity models, then flatten each trained network's parameters into a one-dimensional vector, creating a comprehensive parameter dataset. Next, we employ an autoencoder to learn latent representations of these parameter vectors, capturing essential patterns across diverse PINN's parameters. We then train a conditional diffusion model to store the distribution of these latent vectors, with the corresponding velocity models serving as conditions. Once trained, this diffusion model can generate latent vectors corresponding to new velocity models, which are subsequently decoded by the autoencoder into complete PINN parameters. Experimental results indicate that our method significantly accelerates training and maintains high accuracy across in-distribution and out-of-distribution velocity scenarios.