This short course introduces geoscientists to modern machine learning and deep generative models for solving geophysical inverse problems in mineral exploration.
It begins with an overview of inverse theory and then transitions to state-of-the-art AI methods used to recover complex subsurface structures. Participants will learn the concepts and practical implementation of GANs (generative adversarial networks), cVAEs (conditional variational autoencoder), INNs (invertible neural networks), and NFs (normalizing flows). Hands-on coding exercises and case studies will demonstrate how these models improve geological realism, speed, and uncertainty quantification in inversion workflows.
By the end, attendees will be equipped to integrate deep generative AI into real-world exploration projects.
Duration
Two Days
Who Should Attend?
Students, researchers, and professionals in the mining industry
Level
Intermediate
Prerequisites (Knowledge/Experience/Education required)
- A working knowledge of how deep learning works in general
- An undergraduate-level knowledge of linear algebra, optimization and geophysics
- General knowledge of critical minerals
Learner Outcomes
Students will be able to:
- Implement GANs, INNs and cVAEs for solving geophysical inverse problems in a supervised way
- Solve geophysical inverse problems using normalizing flows in an unsupervised way
- Test geological hypotheses using generative AI models
- Quantify uncertainties of the physical property models from deep learning inversions
