A review of the adjoint-state method for computing the gradient of a functional with geophysical applications

By René-Edouard Plessix. 2006

1,755 citations recorded by the source corpus.

Subfields: Environmental Engineering, Geophysics

View Paper: https://www.semanticscholar.org/paper/07a7a7f05fdbc962f659a903b6d56a695102b63e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Environmental Engineering

In Geophysics

Citations in the corpus

Deep-learning inversion: a next generation seismic velocity-model building method (2019, methodology): “An adjoint-state based gradient descending method was adopted in this experiment (Plessix, 2006).”
Optimal Transport for Seismic Full Waveform Inversion (2016, methodology): “As long as ∇fW 2 2 can be computed efficiently, techniques such as the adjoint state method can be used to efficiently construct the required gradient [14].”
Resolution analysis in full waveform inversion (2011, methodology): “…approach involves a modification of the well-known adjoint method (e.g. Tarantola 1988; Tromp et al. 2005; Fichtner et al. 2006; Liu & Tromp 2006; Plessix 2006; Chen 2011) that allows us to compute H applied to a model perturbation δm, that is,∫ G H(x, y) δm(y) d3y (10) using two forward and two…”
Fast 3D frequency-domain full-waveform inversion with a parallel block low-rank multifrontal direct solver: Application to OBC data from the North Sea (2016, methodology): “INTRODUCTION Full Waveform Inversion (FWI) (Tarantola, 1984) is now routinely used in the oil industry 3 as part of the seismic imaging work-flow, at least in soft geological environments such as the 4 North Sea that make the acoustic parametrization of the subsurface acceptable (see Barnes 5 and Ch”
Full Waveform Inversion and the Truncated Newton Method (2013, methodology): “A survey of its use in seismic imaging has been proposed by Plessix [53].”
Deep-Learning Inversion of Seismic Data (2019, methodology): “Conventional FWI uses gradient-based solvers to update the model parameters, and the gradient is normally calculated through backward waveﬁeld propagation of data residuals based on adjoint-state methods [7], [8].”
Two-dimensional permittivity and conductivity imaging by full waveform inversion of multioffset GPR data: a frequency-domain quasi-Newton approach (2014, methodology): “At each grid point i ∈ [1, NM ] in the medium, the gradient value Gi is computed using the adjoint state method (Plessix 2006) as Gi (m) = Nω∑ Ns∑ Re { uT ∂A ∂mi T v∗ } .”
A new optimization approach for source-encoding full-waveform inversion (2013, methodology): “The computation of the sensitivity matrix ∂u∕∂σ can be avoided by using the adjointstate method (Lions and Magenes, 1972; Lailly, 1983; Tarantola, 1984; Giles et al., 2003; Plessix, 2006; Virieux and Operto, 2009).”
A correlation-based misfit criterion for wave-equation traveltime tomography (2010, methodology): “More details on how to implement the C© 2010 The Authors, GJI, 182, 1383–1394 Journal compilation C© 2010 RAS C© 2010 The Authors, GJI, 182, 1383–1394 Journal compilation C© 2010 RAS adjoint calculations can be found in (e.g. Tromp et al. 2005; Plessix 2006).”
Regularized seismic full waveform inversion with prior model information (2013, methodology): “The data-term gradient is efficiently computed with an adjoint formulation (Plessix, 2006) without an explicit computation of the matrix J.”