Cosine similarity (a.k.a. Salton's cosine) is a measure of structural similarity of two nodes in a network. It counts the number of common neighbours of nodes i and j (given by $n_{ij} = \sum_k A_{ik}A_{kj}$) and divide by the geometric mean of the degrees of i and k:

$\sigma_{ij}= \frac{n_{ij}}{\sqrt{k_i k_j}}$

where $\sigma_{ij}$ is the cosine similarity. The formula is the same as the cosine of the angle between the column of node i and row of node j considered as vectors, hence the name.