Fig. 3: Further analysis of group partitioning by unsupervised learning. | npj Computational Materials

Fig. 3: Further analysis of group partitioning by unsupervised learning.

From: Unsupervised discovery of thin-film photovoltaic materials from unlabeled data

Fig. 3

a A statistical graph of the number of compounds in each group (left), and the ratio of known compounds in each group (right). b Violin plots of Eg of 27 known compounds, ten known compounds in G3, and other 17 known compounds not clustered in G3. The outer shells of the violins bound all data, narrow horizontal lines bound 90% of the data, thick horizontal lines bound 50% of the data, and white dots represent medians. The dashed line shows the position of Eg = 2.0 eV. c Five known I2-II-IV-X4 semiconductors in G3, where I = Ag+. d Five known I2-II-IV-X4 semiconductors in G3, where I = Cu+.

Back to article page