Fig. 1: Description of the bond-type embedded CGCNN (BE-CGCNN) model. | Nature Communications

Fig. 1: Description of the bond-type embedded CGCNN (BE-CGCNN) model.

From: Machine learning-enabled exploration of the electrochemical stability of real-scale metallic nanoparticles

Fig. 1

a Schematic representation of the graph convolution neural network model to predict the adsorption energy. b Representation of bond embedding. Each bond is embedded into a bond vector by one-hot encoding of the bond type. c Optimization of atom embedding features. Costs are compared as a function of various feature combinations. Blue points/line denote the minimum value of each feature combination. Here, GR, AR, EA, and PL indicate the group number, atomic radius, electron affinity, and polarizability, respectively.

Back to article page