SSU computer science majors Ari'El Encarnacion and Cynthia Rosales present research work conducted with Dr. Gurman Gill at COPLAC's Virtual Midwest Regional Undergraduate Research, Scholarly and Creative Activity (URSCA) Conference, held on 14 November 2020. Their presentation "Machine Learning: How Automatic Data Processing can Augment Your Workflow" desceribes a particular application of machine learning to a problem in geology. This work was supported under the NSF project "EarthCube Data Infrastructure: A unified experimental-natural digital data system for analysis of rock microstructures." Their abstract and presentation are below.
Abstract. Applications of Machine Learning include tools that can benefit the sciences, security, consumer algorithms, and even the arts. This study investigates the benefit of machine learning techniques for automatically detecting shear-sense clasts (a type of geological formation, aka sigma clasts) in microscopic images. Machine Learning (ML) is a powerful process that can automatically analyze any type of data and deliver some specified result. Currently, we are using various pre-trained ML models combined with a custom Convolutional Neural Network (CNN), as well as Generative Adversarial Networks (GANs) as tools for classifying clockwise and counterclockwise clasts. Additionally, CNNs require millions of images for training purposes. Since our current image dataset is small (<200 images; roughly < 100 per clast type), we are testing the use of GANs as a tool for image generation.