DuRIN: A Deep-unfolded Sparse Seismic Reflectivity Inversion Network

Abstract

We consider the reflection seismology problem of recovering the locations of interfaces and the amplitudes of reflection coefficients from seismic data, which are vital for estimating the subsurface structure. The reflectivity inversion problem is typically solved using greedy algorithms and iterative techniques. Sparse Bayesian learning framework, and more recently, deep learning techniques have shown the potential of data-driven approaches to solve the problem. In this paper, we propose a weighted minimax-concave penalty-regularized reflectivity inversion formulation and solve it through a modelbased neural network. The network is referred to as deepunfolded reflectivity inversion network (DuRIN). We demonstrate the efficacy of the proposed approach over the benchmark techniques by testing on synthetic 1-D seismic traces and 2-D wedge models and validation with the simulated 2-D Marmousi2 model and real data from the Penobscot 3D survey off the coast of

NuSPAN: A Proximal Average Network for Nonuniform Sparse Model -- Application to Seismic Reflectivity Inversion

“…Such deep neural networks, where the architecture is informed by the inverse problem itself, can provide insights into the problem as well as model interpretability that is critical to gaining physical insights into the system under consideration [44]. In our previous work [47], we demonstrated the efficacy of FirmNet [48] and LISTA-like [28] formulations in solving the seismic reflectivity inversion problem, comparing the approaches with baselines such as BPI [10], [9], FISTA [11], [12], and SBL-EM [46], [16]. Here, we expand the work by constructing and learning from the data [44], a composite nonuniform sparse prior from a convex combination of weighted counterparts of three sparsity-promoting penalties (the 1 norm, MCP [21], and SCAD [23] penalties).…”

Coupling Deep Learning with Full Waveform Inversion

“…In recent years, there are great interest in the FWI community to use deep learning techniques, based on neural networks, to replace the classical least-squares based inversion methods [1,4,16,22,23,28,31,32,36,39,49,54,55,56,57,63,65,66,67,68]. Assume that we are given a set of sampled data…”