Table 1 Local, Quasi-local, and Global Similarity indices.

From: Path-based extensions of local link prediction methods for complex networks

Local index

Matrix form

Global extension

Quasi-local extension

\(CN(u,v) = \left| \Gamma (u)\cap \Gamma (v)\right| \)

\(A^2\)

\(\left( I-\beta A\right) ^{-1}-I\)

\(A^2 + \beta A^3\)

\(RA(u,v) = \sum _{w \in \{ \Gamma (u)\cap \Gamma (v) \}}\frac{1}{ \left| \Gamma (w)\right| }\)

\(AD^{-1}A\)

\(A\left( I-\beta D^{-1}A\right) ^{-1}-A\)

\(AD^{-1}A+\beta AD^{-1}AD^{-1}A\)

\(AA(u,v) = \sum _{w \in \{ \Gamma (u)\cap \Gamma (v) \}}\frac{1}{\log \left| \Gamma (w)\right| }\)

\(A\left( \log D\right) ^{-1}A\)

\(A\left( I-\beta (\log D)^{-1}A\right) ^{-1}-A\)

\(A\left( \log D\right) ^{-1}A+\beta A\left( \log D\right) ^{-1}A\left( \log D\right) ^{-1}A\)

\(SO(u,v) = \frac{2\left| \Gamma (u)\cap \Gamma (v)\right| }{|\Gamma (u)|+|\Gamma (v)|}\)

\(2\left( D\left( A^2\right) _{ij}^{-1}+\left( A^2\right) _{ij}^{-1}D\right) _{ij}^{-1}\)

\(2\left( D\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1}+\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1}D\right) _{ij}^{-1}\)

\(2\left( D\left( A^2+\beta A^3\right) _{ij}^{-1}+\left( A^2+\beta A^3\right) _{ij}^{-1}D\right) _{ij}^{-1}\)

\(SA(u,v) = \frac{\left| \Gamma (u)\cap \Gamma (v)\right| }{\sqrt{|\Gamma (u)|\times |\Gamma (v)|}}\)

\( D^{-\frac{1}{2}}A^2D^{-\frac{1}{2}} \)

\(D^{-\frac{1}{2}}\left( \left( I-\beta A\right) ^{-1}-I\right) D^{-\frac{1}{2}}\)

\(D^{-\frac{1}{2}}A^2D^{-\frac{1}{2}}+\beta D^{-\frac{1}{2}}A^3D^{-\frac{1}{2}}\)

\(LHN(u,v) = \frac{\left| \Gamma (u)\cap \Gamma (v)\right| }{|\Gamma (u)|\times |\Gamma (v)|}\)

\( D^{-1}A^2D^{-1} \)

\(D^{-1}\left( \left( I-\beta A\right) ^{-1}-I\right) D^{-1}\)

\( D^{-1}A^2D^{-1}+\beta D^{-1}A^3D^{-1}\)

\(HP(u,v) = \frac{\left| \Gamma (u)\cap \Gamma (v)\right| }{\min (|\Gamma (u)|, |\Gamma (v)|)} \)

\(\left( \min \left( D\left( A^2\right) _{ij}^{-1},\left( A^2\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

\(\left( \min \left( D\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1},\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

\(\left( \min \left( D\left( A^2+\beta A^3\right) _{ij}^{-1},\left( A^2+\beta A^3\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

\(HD(u,v) = \frac{\left| \Gamma (u)\cap \Gamma (v)\right| }{\max (|\Gamma (u)|, |\Gamma (v)|)} \)

\(\left( \max \left( D\left( A^2\right) _{ij}^{-1},\left( A^2\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

\(\left( \max \left( D\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1},\left( \left( I-\beta A\right) ^{-1}-I\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

\(\left( \max \left( D\left( A^2+\beta A^3\right) _{ij}^{-1},\left( A^2+\beta A^3\right) _{ij}^{-1}D\right) \right) _{ij}^{-1}\)

  1. Here A represents the adjacency matrix of the network, I is the identity matrix with size equal to the size of the matrix A, and D represents the diagonal degree matrix whose ith diagonal element is the degree of the ith node of the graph. Furthermore, \(A^{-1}\) represents the inverse of the matrix A, while \(A^{-1}_{ij}\) represents the element-wise inverse operation.