More graph datasets

This purpose of this repository is to collect and distribute more graph datasets from a variety of sources.

For each graph dataset, we'll have a brief description, along with code to generate that data from the primary source. For the moment, we ask you to please cite

@misc{more-graphs,
  title = "More graph datasets",
  author = "David F. Gleich and Charles Colley and Omar Eldaghar and Meng Liu and Huda Nassar and Haoteng Yin",
  howpublished = "Github: dgleich/more-graphs",
  year = {2019},
  note = "Accessed on",
}      

Other graph repositories

• SuiteSparse Matrix Collection: https://sparse.tamu.edu/
• Colorado Index of Complex Networks: https://icon.colorado.edu/
• Koblenz Network Collection: http://konect.uni-koblenz.de/
• Stanford Network Analysis Project: http://snap.stanford.edu

File formats

The graph and network data is distributed as an SMAT file. The format of the file is:

File: Header Body Header: num_nodes:Int num_nodes:Int num_edges:Int Body: Edge*num_edges Edge: source_id:Int dest_id:Int weight:Double

Undirected edges are listed twice, i.e. both (src,dst) and (dst,src) edges exist.

The node and edge ids run from 0 to num_nodes-1

Example

The following lines describe a 3-node clique in SMAT.

3 3 6
0 1 1
0 2 1
1 2 1
1 0 1
2 0 1
2 1 1

Various other data are stored as bipartite or weighted graphs. Here is an example:

3 4 6
0 0 1.5
1 1 0.5
2 2 1
0 1 0.5
0 4 1
2 1 0.5

Julia code to read the file

Python code to read the file

The Python code takes in a filename, and returns a list of lists representation of the graph.

def read_smat(filename, set_undirected = True):
    """ Load an SMAT file into a list-of-lists graph representation. """
    f = open(filename, 'rU')
    hdr = f.readline()
    parts = hdr.split()
    nverts = int(parts[0])
    ncols = int(parts[1])
    nedges = int(parts[2])
    nremedges = nedges # number of remaining edges
    if nverts != ncols:
        raise ValueError(
            'read_smat line 1: requires nrows (%i) = ncols (%i) for graph'%(
            nverts, ncols))
    graph = [ [] for _ in xrange(nverts) ]
    for lineno, line in enumerate(f):
        parts = line.split()
        if len(parts) == 0: continue
        if nremedges==0: 
            raise ValueError(
                'read_smat line %i: more than %i edges found'%(
                lineno+2, nedges))
        src = int(parts[0])
        dst = int(parts[1])
        if src < 0 or src >= nverts or dst < 0 or dst >= nverts:
            raise ValueError(
                'read_smat line %i: out-of-range edge (%i,%i) found (nverts=%i)'%(
                lineno+2, src, dst, nverts))
        graph[src].append(dst)
        nremedges -= 1
    
    return graph

Matlab code to read the file

The following matlab code reads an SMAT file to a Matlab sparse matrix.

function A = readSMAT(filename)
% READSMAT Load a graph into a Matlab sparse matrix
%   A = readSMAT(filename) where
%   filename is the name of the SMAT file and
%   A is the MATLAB sparse matrix

if ~exist(filename,'file')
    error('readSMAT:fileNotFound', 'Unable to read file %s', filename);
end

s = load(filename,'-ascii');
m = s(1,1);
n = s(1,2);

ind_i = s(2:length(s),1)+1;
ind_j = s(2:length(s),2)+1;
val = s(2:length(s),3);
clear s;
A = sparse(ind_i,ind_j,val, m, n);