Full waveform inversion with compressive updates

Full-waveform inversion relies on large multi-experiment data volumes. While improvements in acquisition and inversion have been extremely successful, the current push for higher quality models reveals fundamental shortcomings handling increasing problem sizes numerically. To address this fundamental issue, we propose a randomized dimensionality-reduction strategy motivated by recent developments in stochastic optimization and compressive sensing. In this formulation conventional Gauss-Newton iterations are replaced by dimensionality-reduced sparse recovery problems with source encodings.