An Introduction to Multilayer Networks
Introduction
Multilayer networks is a general framework to represent the different nations of networks (e.g., monoplex networks, multiplex networks, interdependent networks, networks of networks).
Definitions
A multilayer network is defined as a quadruple:
V is set of nodes
EM⊆VM×VM is the edge set containing the set of pairs of possible combinations of nodes and elementary layers.
L={La}da=1
is the set of elementary layers defined by
d
aspects such that there is one elementary layer set
A node-layer tuple (or simply node-layer) indicates a node
concepts:
- Undirected multilayer network
((u,α),(v,β))∈EM⟹((v,β),(u,α)∈EM) - No self-edges by requiring:
((u,α),(u,α))∉EM - A weighted multilayer network
M
can be defined by assigning weights for the edges in the underlying graph
GM=(VM,EM) (i.e. by mapping each edge of a network to a real number using a function w:EM→R )
Distinguish edges:
- Intra-layer edges
- Inter-layer edges
- Coupling edges
two nodes represent the same entity in different layers:
Multilayer Networks
A. Node-colored Networks
B. Edge-colored Multigraphs
C. Temporal Networks
D.
k
-partite Graphs
Diagnostics for multilayer networks
A. Node degree and neighborhood
Network aggregation
A monoplex network can be constructed by aggregating data from the different layers of a multiplex network and then apply the classic definition of node degree to the resulting monoplex network. That is, the node degree is the number of edges of any type that are incident to a node. However, network aggregation leads to a loss of information and should be avoided if possible.
Therefore it is better to use definitions that maintain the distinction of the layers:
Node Degree
a system composed of
The degree of node
i
on layer
Consequently, the degree of node
i
in a multiplex network is the vector:
Neighborhood
the neighborhood of a node
u
is the set of nodes that can be reached by following any incident edge if
B. Walks, Paths, and Distances
a walk on multilayer networks:
First, if changing layer is considered a step. That is, if there is a cost to change layer.
Second, if there is a difference in taking intra-layer steps in different layers.
Labeled Walks or Compound Relations
Pareto Efficient Path and Pareto Distance
C. Clustering, coefficients, transitivity, and triangles
D. Centrality Measures
Page Rank
Eigenvector Centrality
Katz Centrality
E. Inter-layer Diagnostic
Global Overlap
Global Inter-Clustering Coefficient (ICC)
Degree of Multiplexity
Interdependence
Multiplexity Degree
Dynamics of multilayer networks
mathematical tools: generating functions, spectral theory, tensor algebra
A. Connected Components and Percolation
B. Percolation Cascades
1) start removing a fraction
p
of the nodes uniformly at random
2) divide the remaining nodes into disjoint sets accordingly to the connected component in the first layer
3) update the intra-layer network of the second layer by removing intra-layer edges between nodes that are adjacent to nodes of the first layer which are in a different component
4) updates the intra-layer network of the first layer with the same process used for the second layer
5) repeat 1–4
This process divides the two networks into progressively smaller components until reaching a stationary state in which the nodes in connected components in each of the layers depend only on nodes that are in the same component in the other layer. If one is only interested in a giant component, then we can use a
similar process:
1) remove nodes from the first layer if they are not in that layer’s intra-layer GCC.
2) update the second layer by removing all of the nodes from the second layer that were dependent on the nodes that have been removed from the first layer
3) repeat 1 and 2, but for the second layer
4) repeat 1–3
C. Spreading Models and Diffusion
Empirical multilayer networks
A. Case Study I: Air Transportation Netowork (ATN)
layer: airline company
B. Case Study II: Analysis of Phone Users Sociality
voice call layer
SMS layer
Conclusions
Spreading processes in Multilayer Networks
Introduction
real spreading phenomena are seldom constrained into a single network (called monoplex network). This is evident in online information propagation, where the process of switching network while sharing information on social media has become a basic functionality explicitly provided by many platforms. Another example is represented by the spreading of epidemics propagated by human beings traveling via multiple transport networks (airplanes, trains, etc.).
Three main dimensions for analyzing spreading processes in multilayer networks:
1. how to model the spreading processes
2. what results we can obtain using these models
3. how these results can be exploited in real applications.
Preliminaries
- A monoplex network is a (usually directed) graph.
- A multilayer network is a data structure made of multiple layers, where each layer is a monoplex network.
- A multilayer network where all layers contain exactly the same set of nodes is called multiplex network (partially interconnected multiplex networks);
- each node of a multilayer network belong to exactly one layer, called interconnected (or interdependent) network, in interconnected networks self-interactions across different layers are therefore not possible. Interconnected networks can be viewed as “interconnected communities within a single, larger network”. Examples: power and communication infrastructures.
Modeling spreading processes in multilayer networks
REview and classification of existing spreading models
Epidemic-like models
Goldenberg et al. [85] proposed a discrete-time version of the SIR model called Independent Cascade Model (ICM), where time proceeds in discrete time steps.
In a monoplex network, the probability of transferring an (information) item from one node to another (i.e., transmissibility) is computed as
A recent contribution in the context of multiplex networks [87] proposed a generalized epidemic mean-field (GEMF) model capable of Modeling epidemic-like spreading processes with more complex states in multiplex network layers (compared to two or three states in the SIS and SIR models).
Decision-based models
also called threshold models
Existing decision-based studies follow two different approaches [44]: (i) informational and (ii) direct-benefit effects.
Informational effects
decision is based on the indirect information about the decisions of others. Granovetter presented the first decision-based model, called Linear Threshold Model (LTM)
In LTM, each node chooses a threshold value
TLTM∈[0,1]
and adopts a new behavior if and only if at least a fraction
TLTM
of its neighbors has already adopted the new behavior.
Direct-benefit effects
game-theoretic modeling is at the center of this type of decision-based models.
Theoretical approaches for analyzing spreading models in multilayer networks
Generating function
Markov-chain approximation (MMA)
(i) Discrete-time version [103], and (ii) Continuous-time version
Mean-field theory
Game Theory
Spreading dynamics on multilayer networks
Interconnected networks
Interaction strength between layers
Second-nearest neighbors
Inter-layer pattern
Muitidimensional epidemic threshold
Multiplex networks
Intra-layer structure
Layer similarity
Partially interconnected multiplex networks
Layer-switching cost
spreading velocity
Interacting spreading processes
Diffusion of innovations
Resource constraints
Applications
forward prediction
backward prediction
Influence maximization
Immunization strategies
Epidemic routhing in delay-tolerant networking (DTN)
Malware propagation in the internet
Conclusion and open problems
Empirical study of information diffusion
no works based on real datasets on information diffusion in multilayer networks
graph tomography
Metrics and measurements
New models for spreading processes in multilayer networks
Data visualization
Time-varying networks
Evolution of underlying network structure and spreading process
Outbreak detection