Geophysical model generation with generative adversarial networks

Abstract With the rapid development of deep learning technologies, data-driven methods have become one main research focuses in geophysical inversion. Applications various neural network architectures to inversion seismic, electromagnetic, gravity and other types data confirm potential these real-time parameter estimation without dependence on starting subsurface model. At same time, require large training datasets which are often difficult acquire. In this paper, we present a generator 2D models based generative adversarial networks. Several networks trained separately realistic density stratigraphy reach sufficient degree accuracy generation new highly detailed varied real-time. This allows for creation synthetic cost-effective manner, thus facilitating better algorithms interpretation.