fig4

Figure 4. Comparison of the performance of different uncertainty sampling strategies and ML models on the Au-Ag-Ge system. (A-C) Evolution of Macro-F1 score with increasing training samples; (D-F) Number of labeled samples required by each model to reach a Macro-F1 score of 0.95. US: Uncertainty sampling; ML: machine learning; MLP: multi-layer perceptron; GPC: Gaussian process classifier; SVC: support vector classifier; RF: random forest; XGB: eXtreme gradient boosting.