Multilayer PageRank
Multilayer PageRank (Centrality on Multiplex and Multilayer Networks) · Also known as: multiplex PageRank, layer-coupled PageRank, multilayer random walk centrality, MuxRank
Multilayer PageRank extends the classic PageRank random-walk centrality to networks that contain multiple interconnected layers — such as a social network where people are connected simultaneously via friendship, professional ties, and online platforms. By allowing a virtual walker to jump both within and across layers, the algorithm identifies nodes that are influential across the entire multilayer structure, not just within any single layer.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Multilayer PageRank when your data describe the same set of entities connected through two or more distinct relationship types or contexts (e.g., co-authorship plus citation networks, or multiple social platforms), and you want to identify nodes whose influence extends across layers rather than dominating only one. It is appropriate when layer membership is known, when the network has directed or undirected within-layer links, and when the research question concerns versatility or cross-context centrality. Do NOT use it when you have only a single relationship type (plain PageRank suffices), when layers represent different node sets rather than the same actors across contexts, or when your network is very sparse and small (fewer than ~30 nodes per layer), as random-walk convergence and rank stability become unreliable.
Strengths & limitations
- Captures versatility: identifies nodes influential across multiple relationship contexts simultaneously, which single-layer PageRank cannot detect.
- Handles both directed and undirected within-layer edges and asymmetric interlayer couplings.
- Backed by a rigorous random-walk foundation with guaranteed convergence to a stationary distribution.
- The interlayer coupling parameter allows principled sensitivity analysis to distinguish layer-specific from global influence.
- Scalable via power-iteration solvers to large networks (millions of nodes) when the supra-adjacency is stored in sparse format.
- The choice of interlayer coupling strength is rarely principled and can substantially change the rank order of nodes.
- Building and storing the supra-adjacency matrix has memory cost O(N*L)^2 in dense form, which can be prohibitive for many layers or very large networks.
- Requires that the same set of nodes appears across layers (or a clear mapping), which is not always available in practice.
- Interpretation is less intuitive than single-layer PageRank; explaining versatility scores to non-specialist audiences is challenging.
Frequently asked
How does Multilayer PageRank differ from running PageRank separately on each layer?
Running PageRank per layer gives a score that reflects local influence within that layer only. Multilayer PageRank couples the layers through a supra-adjacency matrix so the random walker can move between layers, producing a single versatility score that reflects cross-layer influence — nodes prominent in only one layer score lower than nodes prominent in many.
How should I choose the interlayer coupling strength?
There is no universal rule. A common practice is to run the algorithm over a range of coupling values and check whether the top-ranked nodes remain stable (rank correlation across values). Domain knowledge — e.g., whether switching social platforms is easy or costly — can also guide the choice. Report sensitivity analyses alongside the main results.
What software implements Multilayer PageRank?
The MuxViz platform (R/Python) by De Domenico directly implements multilayer PageRank and versatility. NetworkX (Python) can be extended to handle supra-adjacency matrices, and igraph provides the underlying eigenvalue solvers used in power iteration.
Is the method appropriate for weighted multilayer networks?
Yes. Within-layer edges can carry weights, which are incorporated into the transition probability matrix before power iteration. Similarly, interlayer coupling weights can differ across layer pairs, allowing finer control over how much each layer-to-layer jump is penalised.
When should I use Multilayer Eigenvector Centrality instead?
Multilayer eigenvector centrality also captures recursive prestige across layers but does not include the damping (teleportation) term of PageRank. It can be unstable on sparse or disconnected layers. PageRank's damping factor makes it more robust in practice; prefer eigenvector centrality only when the theoretical interpretation of purely recursive prestige is important to your research question.
Sources
- De Domenico, M., Sole-Ribalta, A., Omodei, E., Gomez, S., & Arenas, A. (2015). Ranking in interconnected multilayer networks reveals versatile nodes. Nature Communications, 6, 6868. DOI: 10.1038/ncomms7868 ↗
- Boccaletti, S., Bianconi, G., Criado, R., del Genio, C. I., Gomez-Gardenes, J., Romance, M., Sendina-Nadal, I., Wang, Z., & Zanin, M. (2014). The structure and dynamics of multilayer networks. Physics Reports, 544(1), 1–122. DOI: 10.1016/j.physrep.2014.07.001 ↗
How to cite this page
ScholarGate. (2026, June 3). Multilayer PageRank (Centrality on Multiplex and Multilayer Networks). ScholarGate. https://scholargate.app/en/network-analysis/multilayer-pagerank
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Directed PageRankNetwork analysis↔ compare
- Eigenvector CentralityNetwork analysis↔ compare
- Multilayer Betweenness CentralityNetwork analysis↔ compare
- Multilayer Community DetectionNetwork analysis↔ compare
- Multiplex Network AnalysisNetwork analysis↔ compare