Table 1 Overview of the different approaches for approximating \(p_{uv}\)
From: Efficient algorithm to compute Markov transitional probabilities for a desired PageRank
Approach | |
|---|---|
Random | \(p_{uv} \propto 1\) |
Indegree | \(p_{uv} \propto \lvert \{ k \vert (k,v) \in E \}\rvert \) |
Jaccard | \(p_{uv} \propto \frac{\lvert \{ k \vert (u,k) \in E \wedge (v,k) \in E \}\rvert }{\lvert \{ k \vert (u,k) \in E \vee (v,k) \in E \}\rvert }\) |
Traffic | \(p_{uv} \propto \pi _{v}^{*}\) |
(unweighted) PageRank | \(p_{uv} \propto \pi _{v}\) |
ChoiceRank [23] | \(p_{uv} = \frac{\lambda _{v}}{\sum_{(u,v') \in E} \lambda _{v'}}\) (\(\forall \lambda _{i} \in \mathbb{R}_{>0}\)) [24] |
Reverse PageRank | \(p_{uv} \propto \exp (\theta _{uv})\) |