Novel Unsupervised Machine Learning Methods for Data Analytics and Model Diagnostics
V.V. Vesselinov
Machine Learning in Solid Earth Geoscience, Santa Fe2019
Summary
Discussion of Novel Unsupervised Machine Learning Methods for Data Analytics and Model Diagnostics. The presentations addresses why and when unsupervised Machine Learning (ML) should be applied. It covers the use of unsupervised ML for extraction of hidden features characterizing complex datasets and models, and for categorization of data and models based on the extracted features. The presentation also discusses the use of physics-informed ML to improve the efficiency, robustness, and interpretability of ML analyses. Examples from geothermal, geochemical, and climate applications are used to illustrate the concepts and methods.