Simulated petrology data improves mineral predictions from rock chemistry
Simulated petrological labels enable probabilistic mineralogy prediction from whole-rock geochemistry
Machine Learning
Summary
Measuring the minerals inside rocks is important but takes a long time. The authors show that by using computer simulations to create fake mineral data, they can train a machine learning model to guess rock minerals from chemical measurements. Their model works well on volcanic rocks to tell which minerals are present and which are more abundant. They also show a way to adjust predictions to better match real measurements with little extra data. This helps overcome the problem of having too few real labeled samples for training.
What this means in practice
- •For geological survey teams: Estimate mineral abundance in rock samples quickly using chemical data and simulated training instead of slow direct mineral measurements.
- •For mining resource assessors: Identify and rank mineral content in volcanic rock deposits to guide exploration and extraction decisions without full mineralogy assays.
Authors
Matthew J. Liu, Aaron D. Kaplan, Michael L. Whittaker, Kristin A. Persson
Abstract
Modal mineralogy, the mineral identities and abundances in rocks, guides geological studies and industrial resource extraction. While measuring modal mineralogy is time-intensive, whole-rock chemical measurements furnish elemental concentrations at higher throughput. Predicting mineralogy from elemental data, however, poses a one-to-many inverse problem, and labeled mineralogical datasets are too sparse to train on at scale. We address both challenges by simulating petrological labels that express multiple formation conditions per composition and at a scale unattainable by measurements. Nearly one million simulated labels supervise the middle component of a three-part model: a transformer encoder self-supervised on roughly 788,000 whole-rock compositions, the simulation-trained heads that predict distributions over mineral assemblages, and a recalibration layer that aligns the predictions with measured mineralogy. Zero-shot on 640 volcanic rocks, the model discriminates the presence of most mineral categories and predicts which rocks are richer in each mineral. However, the predicted abundances are systematically offset from measured values. Recalibration fit on as few as 25 labeled rocks recovers 76 percent of the likelihood improvement from fitting on the full 368-rock fitting set, while abundance error responds more slowly (11.0% zero-shot and 9.5% with the full fitting set). This work demonstrates how a machine learning model in the geochemical sciences can both leverage an imperfect simulator and mitigate its systematic biases.