Machine Learning in Geosciences
Course Description
Participants will learn the high-level principles of several important topics in machine learning: neural networks, convolutional neural networks, and support vector machine. They will practice the execution of these methods on MATLAB codes (free for 30 days after downloading it from the MATLAB site) and Python-related codes (can be uploaded during the course). Applications include fracture detection in photos, fault delineation in seismic images and picking NMO velocities in semblance gathers.
Course Outline
About 66% of the time will be for 50-minute lectures and the remaining time will be devoted to lab exercises.
Participants’ Profile
The course is designed for geoscientists who have heard about Machine Learning and might know some details, but lack enough knowledge to test ideas or make the next step in understanding. This limitation will be mitigated after a day of diligent attendance and effort. A selective overview of important ML topics is provided and their practical understanding comes from MATLAB and Python-related exercises applied to geoscience problems.
Prerequisites
Participants should have casual familiarity with linear algebra and calculus.
About the Instructor
Gerard T. Schuster received his M.Sc. in 1982 and his Ph.D in 1984 from Columbia University, both in Geophysics. From 1984-1985 he was a postdoctoral fellow at Columbia University, after which he assumed a faculty position in Geophysics at University of Utah from 1985 to 2009. In that time he won several teaching and research awards, founded and directed the UTAM consortium, was chief editor of Geophysics for several years, and supervised more than fifty students to their graduate degrees. He was given EAGE’s Eotvos award in 2007, awarded SEG’s Kauffman gold medal in 2010, and is the 2013 SEG Distinguished Lecturer for spring 2013. In the summer of 2009 he moved to KAUST (King Abdullah University of Science and Technology) as a Professor of Earth Science just north of Jeddah. He holds a joint appointment with both Universities, except he is now an adjunct Professor of Geophysics at University of Utah. His primary interests are in seismic migration and modeling, interferometry, waveform inversion, and a fondness for solving geological problems with modest-sized seismic experiments. Since 2018, he also also been teaching courses on machine learning.