Machine Learning Seismic Attributes Integrations Seismic Inversion for Reservoir Characterization
Machine learning (ML) is part of data analytics. In this book, I emphasize machine
learning. Tom Mitchell, the renowned Carnegie Mellon professor, defined ML as
“the study of computer algorithms that allow computer programs to automatically improve through experience” (1997). He went on to formalize the definition of ML: “A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in .
Where are the Geoscientists in this Digital Ageand ML-Tsunami?
Similar to ML practitioners in many other disciplines, geoscientists are learning
these new ML technologies and terminologies and adopting them for geosciences as needed. This is because many of the concepts widely used in geosciences can be formulated mathematically and analyzed in a better quantitative manner. For example, Markov chains can be used for facies characterization (Krumbein and Dacey 1969; Carr 1982). Krumbein is often cited as the father of mathematical geology.
Krumbein started implementing some fundamental statistical measurements to sedimentology.
Despite the potential for wide-ranging applications, geoscience has been
slow to adopt newML technologies.
There are several reasons behind this, including the seemingly descriptive nature of the discipline, multi-scale heterogeneities, rare events, missing time, challenges in converting some process-based cognitive geologic concepts tomathematical forms, ill-posedness of problems (i.e., more output than the available input), and the lack of labeled large database open to the public. In addition,mmost subsurface data are sampled in a highly biased manner, sparse, and noisy.
The cost of subsurface data acquisition, processing, and maintenance is another obstacle. For example, 3D seismic and downhole fiber-optic data cost millions of dollars for just a small area.
Why should we care about Machine Learning in Geosciences?
Although ML is relatively new to geosciences, we must learn how to apply it to
our problems appropriately. At a high-level, this new tool will enable us to think
more analytically, solve problems consistently and quantitatively, and further transdisciplinary collaboration. ML can help us in several areas of geosciences, including
tectonics, stratigraphy, geophysics, geochemistry, petrophysics, hydrology, paleoclimate, paleontology, remote sensing, and planetary geology. ML will also help us build data-driven models rather than purely conceptual models that are hard to verify with real data.
Web site: www.selftrainingsto.net
E-mail : email@example.com
Mobile : +201120828201,+201010443910
General manager STO(selftrainning Organization)