Dual graph convolutional neural network for predicting chemical networks

BMC Bioinformatics

Table 1 Definitions of several similarity indices between two nodes (G_i,G_j) in a GoG \(\mathcal {G}\)

Similarity index	Definition
Common neighbours	\( \mid \mathcal {N}(G_{i})\cap \mathcal {N}(G_{j})\mid \)
Jaccard’s coefficient	\(\frac {\mid \mathcal {N}(G_{i})\cap \mathcal {N}(G_{j})\mid } {\mid \mathcal {N}(G_{i})\cup \mathcal {N}(G_{j})\mid } \)
Katz	\((I-\beta \boldsymbol {\mathcal {A}})^{-1}-I\)

\(\mathcal {N}(G)\) indicates the neighbor set of node G. I is the identity matrix and \({\mathcal {A}}\) is the adjacency matrix of \(\mathcal {G}\). β is the constant parameter that controls path weights depending on their lengths, and we set β=0.001

ISSN: 1471-2105