Benchmark Network Generation

WebOCD allows you to generate networks and corresponding ground-truth covers using LFR, Signed LFR, and Newman benchmark algorithms. You can e.g. use these benchmarks to evaluate how different algorithms performed relative to the ground truth. You can get to the Benchmarks page by clicking the Benchmarks button on the navbar as shown below.

benchmark-page-button

Here you have the option to choose one of the available benchmark algorithms. You also need to provide the name of the graph (network) and its corresponding cover that will be generated.

benchmark-creation-page

Once you select the algorithm, you’ll be asked to enter the input parameters. These differ between the algorithms. You can also leave the input parameters with the default values. Once you finish the selection, click on the Run Benchmark button at the bottom and this will generate the network and corresponding ground-truth cover.

benchmark-parameter-selection

The results can be viewed on the Networks and Community Detection pages respectively.

Demo

Below you can see a gif of the whole process.

Initially, we’re on the Network Analysis page. From here we click the Benchmarks button in the navbar. This leads to the benchmark algorithm page.

Here we choose one of the available algorithms, in this case, LFR. This opens a page where we need to enter the desired input parameters for LFR. In this case, we choose to set k to 10, n to 500, and on to 50. All other parameters are left with the default values. After selecting the parameters, we choose names for the network (graph) and the cover that will be generated. Clicking on   run-benchmark-btn   runs the algorithm. This redirects us to the _Network Analysis page.

Here, under the Running Benchmarks and Algorithms tab, we can see that the algorithm is running. After several seconds we refresh the page and see that the Running Benchmarks and Algorithms tab is empty. This indicates that the algorithm execution is finished and there are no running algorithms. To view our freshly generated LFR network and cover, we click the Community Detection button which displays the page with all the covers (in this case only one).

benchmark-algorithm-execution