Yunan Yang
Institute for Computational Engineering and Sciences
University of Texas at Austin / Courant Institute
Full waveform inversion (FWI) can be regarded as an iterative technique of solving a wave-equation constrained optimization. We continue to introduce FWI in detail as well as its principal components: the forward modeling, the objective function as a measure of mismatch and the adjoint-state method for fast gradient calculation.
Despite its popularity, FWI is still facing several challenges in realistic field data inversion. These challenges are part of the motivation that inspired us to bring optimal transport and its associated Wasserstein metrics to FWI. Theorems pointing to the advantages of using optimal transport over L2 norm will be discussed, and some large-scale computational examples will be presented. There is an interesting question about how to adapt datasets that are not naturally nonnegative and mass balanced into optimal transport theory.