Fig. 2: Overview of generative learning framework for high-entropy design.
a A generative learning model for the design and discovery of high-entropy dielectric materials. The framework is divided into three steps: (i) generation of the latent space z (ii) classification and sampling of compositions, and (iii) forward inference and inverse design. b Latent space distribution of the different components, purple circles represent the generated 2144 sets of high-performance data, blue squares represent the original experimental data, and solid spheres of different colors represent the five new sets of components predicted by the model. c Entropy versus normalized Ue of candidate materials, where the color of the data points represents their uncertainty.
