Approximate inversion of the wave-equation Hessian via randomized matrix probing

We present a method for approximately inverting the Hessian of full waveform inversion as a dip-dependent and scale-dependent amplitude correction. The terms in the expansion of this correction are determined by least-squares fitting from a handful of applications of the Hessian to random models — a...

Full description

Bibliographic Details
Main Authors: Letourneau, Pierre-David, Demanet, Laurent, Calandra, Henri
Other Authors: Massachusetts Institute of Technology. Earth Resources Laboratory
Format: Technical Report
Language:en_US
Published: Massachusetts Institute of Technology. Earth Resources Laboratory 2014
Subjects:
Online Access:http://hdl.handle.net/1721.1/90470
_version_ 1826218009244467200
author Letourneau, Pierre-David
Demanet, Laurent
Calandra, Henri
author2 Massachusetts Institute of Technology. Earth Resources Laboratory
author_facet Massachusetts Institute of Technology. Earth Resources Laboratory
Letourneau, Pierre-David
Demanet, Laurent
Calandra, Henri
author_sort Letourneau, Pierre-David
collection MIT
description We present a method for approximately inverting the Hessian of full waveform inversion as a dip-dependent and scale-dependent amplitude correction. The terms in the expansion of this correction are determined by least-squares fitting from a handful of applications of the Hessian to random models — a procedure called matrix probing. We show numerical indications that randomness is important for generating a robust preconditioner, i.e., one that works regardless of the model to be corrected. To be successful, matrix probing requires an accurate determination of the nullspace of the Hessian, which we propose to implement as a local dip-dependent mask in curvelet space. Numerical experiments show that the novel preconditioner fits 70% of the inverse Hessian (in Frobenius norm) for the 1-parameter acoustic 2D Marmousi model.
first_indexed 2024-09-23T17:12:41Z
format Technical Report
id mit-1721.1/90470
institution Massachusetts Institute of Technology
language en_US
last_indexed 2024-09-23T17:12:41Z
publishDate 2014
publisher Massachusetts Institute of Technology. Earth Resources Laboratory
record_format dspace
spelling mit-1721.1/904702019-04-12T22:07:56Z Approximate inversion of the wave-equation Hessian via randomized matrix probing Letourneau, Pierre-David Demanet, Laurent Calandra, Henri Massachusetts Institute of Technology. Earth Resources Laboratory Inversion We present a method for approximately inverting the Hessian of full waveform inversion as a dip-dependent and scale-dependent amplitude correction. The terms in the expansion of this correction are determined by least-squares fitting from a handful of applications of the Hessian to random models — a procedure called matrix probing. We show numerical indications that randomness is important for generating a robust preconditioner, i.e., one that works regardless of the model to be corrected. To be successful, matrix probing requires an accurate determination of the nullspace of the Hessian, which we propose to implement as a local dip-dependent mask in curvelet space. Numerical experiments show that the novel preconditioner fits 70% of the inverse Hessian (in Frobenius norm) for the 1-parameter acoustic 2D Marmousi model. National Science Foundation (U.S.); Alfred P. Sloan Foundation 2014-09-30T14:22:03Z 2014-09-30T14:22:03Z 2012 Technical Report http://hdl.handle.net/1721.1/90470 en_US Earth Resources Laboratory Industry Consortia Annual Report;2012-21 application/pdf Massachusetts Institute of Technology. Earth Resources Laboratory
spellingShingle Inversion
Letourneau, Pierre-David
Demanet, Laurent
Calandra, Henri
Approximate inversion of the wave-equation Hessian via randomized matrix probing
title Approximate inversion of the wave-equation Hessian via randomized matrix probing
title_full Approximate inversion of the wave-equation Hessian via randomized matrix probing
title_fullStr Approximate inversion of the wave-equation Hessian via randomized matrix probing
title_full_unstemmed Approximate inversion of the wave-equation Hessian via randomized matrix probing
title_short Approximate inversion of the wave-equation Hessian via randomized matrix probing
title_sort approximate inversion of the wave equation hessian via randomized matrix probing
topic Inversion
url http://hdl.handle.net/1721.1/90470
work_keys_str_mv AT letourneaupierredavid approximateinversionofthewaveequationhessianviarandomizedmatrixprobing
AT demanetlaurent approximateinversionofthewaveequationhessianviarandomizedmatrixprobing
AT calandrahenri approximateinversionofthewaveequationhessianviarandomizedmatrixprobing