Graph and Network algorithms in Julia

First Commit


Last Touched

25 days ago

Commit Count

308 commits


Build Result codecov.io


This package consists of a collection of network algorithms. In short, the major difference between MatrixNetworks.jl and packages like LightGraphs.jl or Graphs.jl is the way graphs are treated.

In LightGraphs.jl, graphs are created through Graph() and DiGraph() which are based on the representation of G as G = (V,E). Similar types exist in Graphs.jl (EdgeList, AdjacencyList, IncidenceList, Graph) - this is again based on viewing a graph G as a set of nodes and edges. Our viewpoint is different.

MatrixNetworks is based on the philosophy that there should be no distinction between a matrix and a network - thus the name.

For example, d,dt,p = bfs(A,1) computes the bfs distance from the node represented by row 1 to all other nodes of the graph with adjacency matrix A. (A can be of type SparseMatrixCSC or MatrixNetwork). This representation can be easier to work with and handle.

The package provides documentation with sample runs for all functions - viewable through Juila’s REPL. These sample runs come with sample data, which makes it easier for users to get started on MatrixNetworks.

Package Installation:

To install package
using MatrixNetworks
To run test cases:

Data available:

For a full list of all datasets:
Loading data example:

Some examples:

largest_component: Return the largest connected component of a graph

Acc is a sparse matrix containing the largest connected piece of a directed graph A p is a logical vector indicating which vertices in A were chosen

A = load_matrix_network("dfs_example")
Acc,p = largest_component(A)
clustercoeffs: Compute undirected clustering coefficients for a graph

cc is the clustering coefficients

A = load_matrix_network("clique-10")
cc = clustercoeffs(MatrixNetwork(A))
bfs: Compute breadth first search distances starting from a node in a graph

d is a vector containing the distances of all nodes from node u (1 in the example below) dt is a vector containing the discover times of all the nodes pred is a vector containing the predecessors of each of the nodes

A = load_matrix_network("bfs_example")
d,dt,pred = bfs(A,1)
scomponents: Compute the strongly connected components of a graph
A = load_matrix_network("cores_example")
sc = scomponents(A)
sc.number #number of connected componenets
sc.sizes #sizes of components
sc.map #the mapping of the graph nodes to their respective connected component
strong_components_map(A) # if you just want the map
sc_enrich = enrich(sc) # produce additional enriched output includes:

Can work on ei,ej:

ei = [1;2;3]
ej = [2;4;1]
bipartite_matching: Return a maximum weight bipartite matching of a graph
ei = [1;2;3]
ej = [3;2;4]
BM = bipartite_matching([10;12;13],ei,ej)
create_sparse(BM) # get the sparse matrix
edge_list(BM)) # get the edgelist
edge_indicator(BM,ei,ej) # get edge indicators