The GraphInsight Software will be presented Tomorrow May,18 at the Startup Spritz, hosted by The Hub Rovereto. The event starts at 6.30pm.

If you like this project, Like our Facebook page.

Below a screenshot of the Internet map.. have fun!

Huge sparse matrices often appear in science or engineering when solving partial differential equations (wikipedia).

Tim Davis from the University of Florida did a great job collecting a myriad of datasets, known as the The University of Florida Sparse Matrix Collection. As the fastcodesign guys say, http://www.fastcodesign.com/1662136/geeky-science-problems-double-as-works-of-art

Timothy Alden Davis, a computer, information science, and engineering professor at the University of Florida, is a sort of Eli Broad with a pocket protector. He’s got the largest open-source collection of sparse matrices in the world [...].

Sparse matrices can be represented as graphs, and this is what they look like using MATLAB/Octave to obtain static visualizations:

How do we map sparse matrices to weighted graphs? the algorithm for square matrices is straightforward: Given a square sparse matrix M = n x n, its corresponding graph G = (N, E) contains n nodes and the edge between nodes n1 and n2 is weighted as M[n1,n2]. If M[n1,n2] = 0, then there is no edge between nodes n1 and n2.

The interested reader may find more details at this link: Visualizing Sparse Matrices. In addition, there are a few academic references at the end of this post.

You can explore this data interactively using GraphInsight with 2D and 3D layouts. To export square sparse matrices from MATLAB/Octave to the DIMACS format (supported by GraphInsight), you can use the following function:

function sparse2dimacs(A,filename)
fid = fopen(filename,'wt');
n=length(A);
for i=1:n
 klen = length( nonzeros(A(i,[1:(i-1) (i+1):end])) );
 fprintf(fid,'%d ( 0.0, 0.0 ) %d ',i, klen );
 % find non zero elements of row i
 [p]=find(A(i,:));
 for m=p
  if (i~=m )
   fprintf(fid,'%d ',m);
   end
  end
  fprintf(fid,'\n');
end

In the following, we show some screenshots obtained from GraphInsight. We observe that these are screenshots from interactive 3D visualizations.

Explore interactively your data with GrahInsight! download your evaluation copy at graphinsight.com. We support Microsoft Windows, Mac OS X, and Linux. Do not hesitate to contact us for any question or problem you may have with our product.

We are seeking for collaborations and work opportunities. Do not hesitate to contact Us at info@graphinsight.com.

Bibliographic notes

We loaded the data in GraphInsight, and this is the result.

Explore your data with GrahInsight! download your evaluation copy at graphinsight.com. We support Microsoft Windows, Mac OS X, and Linux.

“I find that the 3rd dimension is really helpful to find central nodes which would otherwise be “flattened” and hidden in a 2D structure.”, cit.

Despite the clear benefits of 3D in some visualization problems, the time required to determine the 2D and 3D layouts is similar: 2D embeddings have one less degree of freedom if compared to 3D embeddings, but 3D methods can unravel the network more easily because of the better “spatial” freedom.

To appreciate 3D network models you need an interactive application: without interaction 3D does not really exist. The GraphInsight software is a solution to this problem: Load your network, pick a layout technique, and interact with your data in 2D and 3D.

PS. Did you know that with GraphInsight you can mix 2D and 3D? Three simple steps: (i) generate a layout for your data with your preferred 2D method, (ii) switch to the “Nothing” method in 3D, (iii) press the “Start” button again. You did it! The graph is flattened onto a 2D plane and you can rotate it in a 3D space.

Download your free evaluation copy of GraphInsight at graphinsight.com. We support Microsoft Windows, Mac OS X, and Linux.

[1] “Science overlay maps: a new tool for research policy and library management” by Rafols, I. and Porter, A.L. and Leydesdorff, L. - Journal of the American Society for information Science and Technology.

Thank You for the feedback, your advices are our TODO list

We are seeking for collaborations and work opportunities. Do not hesitate to contact Us at info@graphinsight.com.

Have fun!

GraphInsight has also been accepted on CrunchBase: GraphInsight CrunchBase profile.

I wrote this software because I felt frustrated by other similar tools such as Graphviz and Gephi. GraphViz is great but not interactive, and Gephi is slow and unstable. I really needed a better tool, and this is my answer to the problem.

This software would not be here without the help of Carlo Nicolini: He did a great job with graph drawing algorithms. We got it right this time!

I would thank Cristian Stenico for our nice logo, and Aaron Zatelli, Paolo Cerchiaro for the many interesting discussions and passionate support.

GraphInsight has been accepted on CrunchBase, we have our own profile now: GraphInsight CrunchBase profile. Follow us.