CodePlexProject Hosting for Open Source Software

This page provides an index to the materials used in the two studies described in the paper:
*Edge Compression Techniques for Visualization of Dense Directed Graphs*, by Tim Dwyer, Nathalie Henry Riche, Kim Marriott and Chris Mears.

**Study 1**

The graph files are available here in DGML format (an XML format viewable in Visual Studio 2012 Pro or Ultimate). There are nine distinct graphs in three levels of difficulty (7, 10 or 15 nodes).

The labels for each of these nine graphs were permuted in two different ways to obtain 18 flat graphs.

Each of these 18 graphs were presented to users using one of three techniques: Flat (no compression), Simple Matching, or Modular Decomposition. The 18*3=52 images are here.

Also, there are 18 training images, obtained from different graphs, that can be found in three directories named "Training" under each technique directory above.

**Study 2**

The flat graphs and the sets of modules generated by our MiniZinc model are here.

Each file is an optimal solution for one of three edgeweight and crossingweight parameter settings obtained using the CPX solver that comes with the MiniZinc distribution.

The file names indicate:

The actual images used in the study are here. The file names indicate:

**Compressed Graph Corpus**

The power-law generated graphs described in Section 6 are available here.

There are 20 graphs and multiple files for each graph. The .dzn file is the simple adjacency matrix we use as input to the solver. The .out-X-Y files are the solutions obtained for edgeweight=X, crossingweight=Y. Each solution file contains all the intermediate solutions found by the solver. The final optimal solution is the last one.

The graph files are available here in DGML format (an XML format viewable in Visual Studio 2012 Pro or Ultimate). There are nine distinct graphs in three levels of difficulty (7, 10 or 15 nodes).

The labels for each of these nine graphs were permuted in two different ways to obtain 18 flat graphs.

Each of these 18 graphs were presented to users using one of three techniques: Flat (no compression), Simple Matching, or Modular Decomposition. The 18*3=52 images are here.

Also, there are 18 training images, obtained from different graphs, that can be found in three directories named "Training" under each technique directory above.

The flat graphs and the sets of modules generated by our MiniZinc model are here.

Each file is an optimal solution for one of three edgeweight and crossingweight parameter settings obtained using the CPX solver that comes with the MiniZinc distribution.

The file names indicate:

- the flat graph used as described in the paper (e.g. Difficult1NonModular)
- the optimization parameters used (edgeweight and crossingweight) to generate the decompositions

The actual images used in the study are here. The file names indicate:

- the flat graph used as described in the paper (e.g. Difficult1NonModular)
- the optimization parameters used (edgeweight and crossingweight) to generate the decompositions
- the task number: corresponding to different "start" and "end" nodes and randomized labelling

The power-law generated graphs described in Section 6 are available here.

There are 20 graphs and multiple files for each graph. The .dzn file is the simple adjacency matrix we use as input to the solver. The .out-X-Y files are the solutions obtained for edgeweight=X, crossingweight=Y. Each solution file contains all the intermediate solutions found by the solver. The final optimal solution is the last one.

Last edited Jul 3, 2013 at 2:44 AM by tgdwyer, version 6