Feature-based Inversion of 2.5D Controlled Source Electromagnetic Data using Generative Priors

TL;DR

This paper presents a feature-based 2.5D electromagnetic data inversion method that leverages generative priors via a variational autoencoder, improving reconstruction accuracy and flexibility over traditional approaches.

Contribution

It introduces a novel plug-and-play inversion framework using a VAE to incorporate prior information without black-box neural network mappings.

Findings

Effective incorporation of prior information improves accuracy

Framework demonstrates good generalization on field data

Flexible adaptation to different survey configurations

Abstract

In this study, we investigate feature-based 2.5D controlled source marine electromagnetic (mCSEM) data inversion using generative priors. Two-and-half dimensional modeling using finite difference method (FDM) is adopted to compute the response of horizontal electric dipole (HED) excitation. Rather than using a neural network to approximate the entire inverse mapping in a black-box manner, we adopt a plug-andplay strategy in which a variational autoencoder (VAE) is used solely to learn prior information on conductivity distributions. During the inversion process, the conductivity model is iteratively updated using the Gauss Newton method, while the model space is constrained by projections onto the learned VAE decoder. This framework preserves explicit control over data misfit and enables flexible adaptation to different survey configurations. Numerical and field experiments demonstrate that the proposed approach effectively incorporates prior information, improves reconstruction accuracy, and exhibits good generalization performance.