Machine learning inversion of geophysical data by a conditional variational autoencoder

Recovering geologically realistic physical property models by geophysical inversion is a long-standing challenge. Generative neural networks offer a promising path to meeting this challenge because they can produce spatially complex models that exhibit the characteristics of a set of training models, even when those characteristics are not easy to quantify. Here we develop a framework for incorporating prior geological knowledge into geophysical inversions using conditional variational autoencoders (CVAE). Once trained, the decoder network of the CVAE inverts the data by taking as input the observed data and a set of latent features which are sampled from a standard normal distribution. CVAE inversion reproduces the observed data and incorporates the information embedded in the training models.