Augmented flexible Krylov subspace methods with applications to Bayesian inverse problems (Q6966778)

This interesting paper studies augmented flexible Krylov subspace methods with applications to Bayesian inverse problems. More specifically, the paper studies linear ill-posed inverse problems of the form \(Bu_{\mathrm{true}}+e=b\), \(B\in \mathbb R^{m\times n}\) where \(e\sim N(0,r)\) is some unknown noise with a symmetric positive definite covariance \(r\), \(b\) is a measurement which depends on \(r\) and what is utrue is the unknown object the authors wish to reconstruct. The singular values of \(B\) decay and cluster at zero without an evident gap between two consecutive ones.\N\NMore specifically, the authors are interested in solutions of \(u_{\mathrm{true}}=x+\eta\) that can be modeled as the sum of two independent random variables where \(x\) is smooth and \(\eta\) is sparse. The paper presents two new augmented flexible Krylov subspace methods, AF-LSQR and AF-GMRES to calculate solutions to these large-scale linear discrete ill-posed problems. The efficiency of the new methods is shown through a variety of experiments for example including a synthetic image deblurring problem, a synthetic atmospheric transport problem, and fluorescence molecular tomography reconstructions using both synthetic and real world experimental data. \N\NThe paper is well written with a very good set of references.