Fig. 1: Distribution of training and validation set samples in the feature space. | npj Computational Materials

Fig. 1: Distribution of training and validation set samples in the feature space.

From: Graph attention networks decode conductive network mechanism and accelerate design of polymer nanocomposites

Fig. 1

a–h Carbon nanotube concentrations ranging from 1% to 8%. The spatial comparison between concentrations is performed using SVD and t-SNE dimensionality reduction. The distributions of the training and validation sets for CP1–CP8 are uniform and representative, which ensures the reliability of the model evaluation.

Back to article page