Full waveform inversion with random shot selection using adaptive gradient descent

TL;DR

This paper introduces a novel mini-batch full waveform inversion method using the Adam optimizer with adaptive gradient descent and random shot selection, reducing computational costs while maintaining high-resolution subsurface models.

Contribution

It presents a new approach combining mini-batch FWI with Adam optimization, including criteria for hyperparameter selection and shot batch size, demonstrated on synthetic Marmousi data.

Findings

Reduces computational cost of FWI using mini-batches

Provides stable model updates with Adam optimizer

Effective on synthetic Marmousi model data

Abstract

Full waveform inversion (FWI) is a powerful yet computationally expensive technique that can yield subsurface models at high resolution. Randomly selected shots ("mini-batches") can be used to approximate the misfit and the gradient of FWI, thereby reducing its computational cost. Here, we present a methodology to perform mini-batch FWI using the Adam algorithm, an adaptive optimization scheme based on stochastic gradient descent. It provides for stable model updates by smoothing the gradient across iterations and can also account for the curvature of the optimization landscape. We describe empirical criteria to choose the hyperparameters of the Adam algorithm and the optimal mini-batch size. The performance of the outlined scheme is illustrated on synthetic data from the Marmousi model.