Fig. 11: Comparison of model performance and computational complexity under different training strategies. | npj Computational Materials

Fig. 11: Comparison of model performance and computational complexity under different training strategies.

From: A crystal graph convolutional neural network framework for predicting stacking fault energy in concentrated alloys

Fig. 11

a, b Mean absolute error (MAE) on the test set as a function of the number of training samples for Dataset 1 and the stacking fault (SF) dataset, respectively. c The computational complexity of the hierarchical training strategy compared with the training strategy used in Fig. 4.

Back to article page