In a [[Network Types#^d09159|directed network]] you can have two kinds of special nodes:
- *Sink node:* Only has incoming edges but no outgoing edges (i.e. out-degree is zero)
- *Source node:* Only has outgoing edges but no incoming edges (i.e. in-degree is zero)
When a [[Network Types#^7075f3|Walk]] passes such a kind of node, it will be stuck forever. This is problematic for [[Eigenvector Centrality]], as the sink / source node has centrality $0$, and this will be propagated through the whole network.
We can circumvent this effect by giving every node at least some centrality $\beta$ for free.
$
\begin{align}
x_i^{(k)} &= \alpha \sum_{j=1}^n A_{ij}x_j^{(k-1)} + \beta \tag{1}\\[10pt]
x^{(k)}&= \alpha Ax^{(k-1)} + \beta \tag{2}\\[8pt]
x^{(k)} &=\alpha A^kx^{(0)}+ \beta \tag{3}
\end{align}
$
As $k$ goes to infinity the $x^{(k)}$ converges to a vector $v$, which is a generalization of the eigenvectors.
$ x^{(k)}\xrightarrow[]{k \to \infty}v; \quad \quad v=(\mathbf I- \alpha A)^{-1}\beta $