Robust inversion via semistochastic dimensionality reduction

Abstract

We consider a class of inverse problems where it is possible to aggregate the results of multiple experiments. This class includes problems where the forward model is the solution operator to linear ODEs or PDEs. The tremendous size of such problems motivates dimensionality reduction techniques based on randomly mixing experiments. These techniques break down, however, when robust data-fitting formulations are used, which are essential in cases of missing data, unusually large errors, and systematic features in the data unexplained by the forward model. We survey robust methods within a statistical framework, and propose a semistochastic optimization approach that allows dimensionality reduction. The efficacy of the methods are demonstrated for a large-scale seismic inverse problem using the robust Student's t-distribution, where a useful synthetic velocity model is recovered in the extreme scenario of 60% data missing at random. The semistochastic approach achieves this recovery using 20% of the effort required by a direct robust approach.

Mathematical Programming, 2012

Cited by in corpus (19)

• Stop Wasting My Gradients: Practical SVRG
• Estimating Nuisance Parameters in Inverse Problems
• Sub-Sampled Newton Methods I: Globally Convergent Algorithms
• Sub-Sampled Newton Methods II: Local Convergence Rates
• Newton-Type Methods for Non-Convex Optimization Under Inexact Hessian Information
• Variational properties of value functions
• Large-scale Log-determinant Computation through Stochastic Chebyshev Expansions
• Convergence Rates for Inverse Problems with Impulsive Noise
• Approximating the Spectral Sums of Large-scale Matrices using Chebyshev Approximations
• A Data-Scalable Randomized Misfit Approach for Solving Large-Scale PDE-Constrained Inverse Problems
• Generalized Kalman Smoothing: Modeling and Algorithms
• Target-oriented full-waveform inversion based on generalized Rényi entropy using patched Green's function techniques
• Minimum Distance Estimation for Robust High-Dimensional Regression
• A Semismooth Newton Stochastic Proximal Point Algorithm with Variance Reduction
• Error bounds for some approximate posterior measures in Bayesian inference
• l1-Norm Minimization with Regula Falsi Type Root Finding Methods
• Robust model training and generalisation with Studentising flows
• XConv: Low-memory stochastic backpropagation for convolutional layers
• General Optimization Framework for Robust and Regularized 3D Full Waveform Inversion