Anderson acceleration for seismic inversion

State-of-the-art seismic imaging techniques treat inversion tasks such as full-waveform (FWI) and least-squares reverse time migration (LSRTM) partial differential equation-constrained optimization problems. Due to the large-scale nature, gradient-based algorithms are preferred in practice update model iteratively. Higher-order methods converge fewer iterations but often require higher computational costs, more line-search steps, bigger memory storage. A balance among these aspects has be considered. We have conducted an evaluation using Anderson acceleration (AA), a popular strategy speed up convergence of fixed-point iterations, accelerate steepest-descent algorithm, which we innovatively iteration. Independent unknown parameter dimensionality, cost implementing method can reduced extremely low dimensional problem. The further by low-rank update. determine theoretical connections differences between AA other well-known L-BFGS restarted generalized minimal residual compare their requirements. Numerical examples FWI LSRTM applied Marmousi benchmark demonstrate effects AA. Compared with method, achieve faster provide competitive results some quasi-Newton methods, making it attractive for inversion.