Fig. 1: Overview of the CIGNN model architecture. | npj Computational Materials

Fig. 1: Overview of the CIGNN model architecture.

From: Charge integrated graph neural network-based machine learning potential for amorphous and non-stoichiometric hafnium oxide

Fig. 1: Overview of the CIGNN model architecture.

a Graph representations used in the CIGNN model, GAB and GBA. Red represents nodes, and blue represents edges in each graph. b CIGNN model components: CPM and EFPM. Gray, green, red, and sky blue boxes represent the inputs, embedding layers, graph convolution layers, and outputs, respectively, in both CPM and EFPM. The yellow box represents the QEq layer in CPM. The light blue box represents the element scale layer in EFPM.

Back to article page