Fig. 3: Comparative performance metrics for crystallographic symmetry classification using DL Models. | npj Computational Materials

Fig. 3: Comparative performance metrics for crystallographic symmetry classification using DL Models.

From: Deep learning for symmetry classification using sparse 3D electron density data for inorganic compounds

Fig. 3

This figure presents bar plots that compare the performance of coGN, FCN, PointNet, and Sparse 3D CNN models in three symmetry classification categories: crystal system (left), extinction group (middle), and space group (right). For each category and model, the broader bars represent top-1 hold-out test accuracy, while the overlaid lighter bars indicate the F1 score, a measure of test precision. a Displays hold-out test accuracy results obtained using the 120k_MP dataset. b Shows hold-out test accuracy outcomes from the 190k_ICSD dataset. Note: The results for the coGN model with the 190k_ICSD dataset are excluded due to compatibility issues between the model and the dataset.

Back to article page