Fig. 5: Uniqueness and novelty check of the generated materials. | npj Computational Materials

Fig. 5: Uniqueness and novelty check of the generated materials.

From: Generative adversarial networks (GAN) based efficient sampling of chemical composition space for inverse design of inorganic materials

Fig. 5

a Comparison of uniqueness curves of the hypothetical materials generated by three GANs. GAN-MP achieves the dominating curve due to its more balanced distribution of binary/ternary/quaternary training samples. b Distribution of recovery rates of training and validation samples, and also percentages of new generated hypothetical materials.

Back to article page