Regularization of anisotropic full-waveform inversion with multiple parameters by adversarial neural networks

The anisotropic full-waveform inversion (FWI) is a seismic inverse problem for multiple parameters, which aims to simultaneously reconstruct the vertical velocity and parameters of earth’s subsurface. This multiparameter suffers from two issues. First, objective function data fitting less sensitive parameters. Second, crosstalk effect among different worsens model update in iterative inversion. We have developed method that statistically regularizes FWI using Wasserstein adversarial networks, by penalizing distance between distribution current at borehole locations. regularizer can mitigate issues with therefore it also be applied other problems