Energy and Geoscience Institute

The short course Machine Learning Methods in Science consists of lectures that explain the high-level concepts underlying fully connected neural networks, convolutional neural networks (CNN), clustering methods, Recurrent Neural Networks, and ChatGPT. It is intended for practicing scientists and engineers in geosciences and engineering who have little knowledge of ML methods but desire a hands-on introduction to Machine Learning (ML) methods and their applications to scientific problems. The goal is to learn the basics of some ML methods and their implementation on free software that can be used for solving real-world problems. Computer exercises consist of cluster identification of geochemical signatures for rock formations associated with hominin fossils near Lake Turkana, CNN identification of bird types from photographs, CNN detection of cracks along a fractured dam face from drone photos, and discrimination of destructive weeds from plants in drone photos near Kennecott mine.