On the role of memorization in learned priors for geophysical inverse problems

Ali Siahkoohi, Davide Sabeddu

Learned priors based on deep generative models offer data-driven regularization for seismic inversion, but training them requires a dataset of representative subsurface models -- a…