Dallas Geophysical Society Luncheon - January 2020
Date: Tuesday, January 14, 2020
Time: 11:30 am - 1:00 pm
Topic: Attribute selection and supervised seismic facies classification using Probabilistic Neural Networks
Speaker: David Lubo-Robles
Abstract: Machine learning algorithms such as Artificial Neural Networks (ANN), and Self-organizing Maps (SOM) have been adopted by geoscientists to accelerate the interpretation of their data, and achieve higher levels of accuracy during reservoir characterization. Identifying the best combination of attributes needed to perform either supervised or unsupervised machine learning tasks continues to be the most-asked question by interpreters. Stepwise regression is the most well-established technique for selecting the best number and combination of attributes to be used. Unfortunately, it does not test all the possible combinations of seismic attributes; thus, missing important relationships existing between them. In this study, we develop a novel technique called Exhaustive PNN which couples the probabilistic neural network’s capacity in exploring non-linear relationships with an exhaustive search algorithm in order to determine the optimal combination of seismic attributes to distinguish between salt and non-salt seismic facies in a Gulf of Mexico 3D seismic survey.
Bio: David Lubo-Robles received a bachelor’s degree in geophysical engineering from Simon Bolivar University, Venezuela, and a master’s degree in geophysics from the University of Oklahoma under Kurt J. Marfurt. He is a student member of the Society of Exploration Geophysicist and currently is pursuing a doctorate in geophysics at the University of Oklahoma, studying under Kurt J. Marfurt and Matthew Pranter. His research interests include the development and application of modern machine learning and pattern recognition techniques, together with quantitative interpretation skills, including pre-stack inversion and seismic attribute analysis to delineate geologic features amenable to hydrocarbon accumulation.