Machine learning for mineral exploration: a data odyssey

Abstract:
Remote sensing enables us to observe our planet using different types of data representation with guidance from satellites, airplanes, and drones. It has always been challenging to process remote sensing data due to computational complexities for detecting features of interest, such as lithological units, hydrothermal alteration zones, and geological structures mainly caused by noise and sparse information. However, there has been good progress in developing machine learning methods to facilitate processing and interpreting remote sensing data in the last decade. In this seminar, we present innovations with advancements in machine learning for creating geoscientific models and processing remote sensing data to unravel the geological and climate history of the planet and to aid in prospecting for mineral resources. We first present a framework that couples machine learning models with plate tectonics that unravel the formation of porphyry copper deposits for the past 80 million years in the western edge of America and highlight effective factors in mineralization. We then present a deep learning-based approach using convolutional neural networks to process remote sensing data and create lithological maps. Then we give an update on ongoing projects in the scope of using deep autoencoders and remote sensing data for lithological mapping and data augmentation for addressing limited and imbalanced training samples in mineral exploration problems.
Download pdf slides: github.com/rohitash-chandra/r...
About the Presenters:
Dr. Rohitash Chandra is a Senior Lecturer in Data Science at the UNSW School of Mathematics and Statistics. Dr. Chandra leads a program of research encircling methodologies and applications of artificial intelligence, particularly in Bayesian deep learning, neuro-evolution, climate extremes, geoscientific models, and mineral exploration. Dr. Chandra has developed novel methods for machine learning inspired by neural systems and learning behaviour that include transfer and multi-task learning, with the goal of modular deep learning. His current interests are ensemble learning, data augmentation, applied language models, bioinformatics, and machine learning for COVID-19. research.unsw.edu.au/people/d...
Dr. Ehsan Farahbakhsh is a Research Associate in the EarthByte Group, School of Geosciences, University of Sydney. He holds a PhD in Mining Engineering - Mineral Exploration and is currently involved in several projects in the scope of developing data science applications, particularly machine learning and deep learning in mineral exploration. Along with his research activities, he has been involved in several projects as an exploration geologist or spatial data analyst for the exploration industry, primarily for providing prospectivity maps of various ore deposit types from regional to deposit scale. His research interests are multidimensional mineral prospectivity modelling, geological remote sensing, geostatistics, and the application of data science and UAVs in mineral exploration. www.sydney.edu.au/science/abo...

[@agentstona]

Good presentation just one feed back . A validation dataset is a sample of data held back from training your model that is used to give an estimate of model skill while tuning model’s hyperparameters. The validation dataset is different from the test dataset that is also held back from the training of the model, but is instead used to give an unbiased estimate of the skill of the final tuned model when comparing or selecting between final models. This is why it is commonly expected to show validation results as well in presentations no matter how similar the data set is it is standardization , a check list item and one of the things that one is always expected to do and show case when presenting AI related projects..

[@Actor-Vishharad]

Good work