Paper Summary
Machine learning dominated the technical program at the Society of Exploration Geophysicists (SEG) conference this year; with Deep Learning applications being most popular. A consideration of the mechanics of convolutional neural networks (CNNs) and generative adversarial networks (GANs) leads to a comparison of several complementary efforts to resolve the most obvious challenge to such pursuits in the geosciences (the lack of real training data), and several complementary efforts to resolve one of the key weaknesses in Full Waveform Inversion (the lack of very low frequency signal in the recorded data). Despite the hype that can accompany this broad topic, encouraging progress is being made towards geoscientists being able to make better informed decisions, using more (all) data, and in less time.