NeuroData's Graph DataBase

Lowering the barrier to entry for scalable neuroscience

Specifications for acceptable input graphs

• Any igraph supported format is the preferred format for data ingestion.
• A MATLAB struct with a single data element that is of type numpy CSC (Compressed Sparse Column) matrix.
• Any 2 Dimensional square MATLAB matrix.

It should be loadable from the python shell via the following call: scipy.io.loadmat(filename)['your_data_element_name'].
Such a matrix can be created by creating an MATLAB matrix and using the save function or via Python's scipy.io.savemat function.

Single graph

As long as your graph meets one of the above specifications you may compute invariants on any graph of your choosing.

Programmatic use

This is can be accessed using the base url

http://openconnecto.me/graph-services/graphupload/

The example python script can be found here.

How to use the example

After downlaoding the example above, start a terminal session. Assuming you are in the same directory as the example. Typing python compute.py -h will give help on how to use the example and what options are available to you.

Single graph

Assume we have a single graph located in the same directory as the example script named GRAPH_FILE. Typing

python compute.py http://openconnecto.me/graph-services/graph-services/graphupload \
GRAPH_FILE email@domain GRAPH_FORMAT \
-i cc tri ss1

will upload your GRAPH_FILE file, in GRAPH_FORMAT format; compute: clustering coefficient, local triangle count, and scan statistic-1.

Here is a breakdown of the command line call arguments:

• http://openconnecto.me/graph-services/graphupload/

  is the base url for uploading graphs for invariants to be computed.
• cc,tri,ss1 is a comma separated (no spaces) list of the invariants you want computed.
  The only available options for invariants are:

  • ss1 for the Scan Statistic 1.
  • tri for the Triangle Count.
  • cc for the Clustering Coefficient.
  • mad for the Maximum Average Degree.
  • deg for The Local Degree Count.
  • eig for the Top-k (100 or maximum available/computable) eigenvalues and eigenvectors.
• The second argument allows you to pass the input format of the graph(s) being uploaded. Current available options for graph format are:
  graphml, ncol, edgelist, lgl, pajek, graphdb, mat, npy

Starting a tmux or screen session is recommended as you must wait for the http response to otbain a url containing the result.