Machine learning in pre- and post-imaging seismic data with different roles of wave physics

Tuesday, October 23, 2018 - 10:05am - 10:55am
Keller 3-180
Weichang Li (Aramco Services Company)
For machine learning, the luxury of having wave physics models can be confusing sometimes. In this talk I will present two machine learning examples involving prestack and imaged seismic data, respectively, and show different ways that wave physics can be incorporated to help with feature representation and improve learning performance. Specifically, the first example predicts facie types from pre-stack seismic gathers, aided by well-log records. The second example detects and classifies geobody structures from seismic cross sections.