Fig. 2
From: Interpretable machine learning models classify minerals via spectroscopy
The final trained classifiers for recognizing secondary oxyanion and charge balancing species in uranium minerals from our methodology are presented for top row, left to right: phosphates, vanadates, and silicates, and bottom row, left to right: water, copper, and a visual representation of how these models combine to provide information on a mineral arsenate. For each model, blue dots represent training points that belong to the class while red dots represent training points that do not. The background color represents the model’s confidence that a point is (blue) or is not (red) a member of that class, with darker shades representing a higher confidence. The axes reflect the correspondence of spectral peak positions identified by the classifier training scheme, as outlined in the Methods17. Lower right – pedagogical depiction of classification contribution to overall mineral assignment.
