Matt Stabeler

In Plotting all parameters to GCE-H in BubbleH simulation I showed the results of plotting lots of parameters to GCEH, in one simulation run. Here are some more details:

The process worked as follows:

1. Generate edge list file with weights for run period
2. Run GCEH script with multiple parameter values, on the edge list
3. Generate community list files, and hierarchy files from each output of GCE for use by LocationSim
4. Prepare configuration files for community ranking and simulation
5. Generate community rankings for each community
6. Run simulation for each community set
7. Visualise output

1) Edge List

 java -jar dtnsim.jar xml/graphs/edgelist.xml 4 DATASET=mit-nov-cheat

2) Run GCEG Script

mattstabeler@erdos:~/GCE/master$ php -f ~/LocationSim/scripts/clustering/batchGCE.php ~/LocationSim/OUTPUT/edgelist-graphs/mit-nov-cheat/edge_list.dat

 ' . $outfile  . '.gce_output_K-${K}_ST-${ST}_MAP-${MAP}_E-${E}_Z-${Z}.dat';

  foreach($ks as $k){
   foreach($es as $e){
     foreach($sts as $st){
       foreach($maps as $map){
          foreach($zs as $z){
            $run = preg_replace('/${K}/', $k, $command);
            $run = preg_replace('/${E}/', $e, $run);
            $run = preg_replace('/${Z}/', $z, $run);
            $run = preg_replace('/${MAP}/', $map, $run);
            $run = preg_replace('/${ST}/', $st, $run);
            $data = passthru($run);
            //~ `$run`;
          }
        }
      }
    }
  }

}
?>

3) Generate files for LocationSim

mattstabeler@erdos:~/LocationSim/scripts/transform$ php -f ConvertGCEHOutput.php ~/LocationSim/OUTPUT/edgelist-graphs/mit-nov-cheat/edge_list.dat.gce_output*.dat

This generates: a re-written gce output file, which has consecutively numbered community ID’s from 0 to N (edge_list.dat.gce_output_K-3_ST-0.5_MAP-0.5_E-0.15_Z-0.2.dat.renumbered.dat), a LocationSim compatible list of communitites (edge_list.dat.gce_output_K-3_ST-0.5_MAP-0.5_E-0.15_Z-0.2.dat.communites.dat) and a JSON representation of the communities for use in LocationSim (edge_list.dat.gce_output_K-5_ST-0.9_MAP-0.9_E-0.25_Z-0.2.dat.json).

toJson());
  file_put_contents($out_file . '.communites.dat', $parser->toSimCommunityList());
  file_put_contents($out_file . '.renumbered.dat', $parser->toString());
}
?>

Move the output of all files to /datasets/communities/{dataset namet}/{community type}
e.g /home/mattstabeler/LocationSim/datasets/communities/mit-nov-cheat/GCEH/

4) Config files
Datasets config:

~/LocationSim/xml/datasets/gceh-communities.xml

Community Generation Config

~/LocationSim/xml/bubbleH/centrality-k.xml

Simulation Config

~/LocationSim/xml/bubbleH/bubbleH-k.xml

5) Generate community rankings

 mattstabeler@erdos:~/LocationSim$ java -jar dtnsim.jar xml/bubbleH/centrality-k.xml 10

6) Run simulation

 mattstabeler@erdos:~/LocationSim$ java -jar dtnsim.jar xml/bubbleH/bubbleH-k.xml 10

During the run, the system used only 4 cores (out of 24) and 6.1GB of memory (out of 124GB) – and should not have affected other users considerably (hopefully!).

top command at runtime

7) Visualise output

mattstabeler@erdos:~/LocationSim$ python scripts/plot/plot-barchart-special.py 2 2 OUTPUT/bubbleH/mit-nov-cheat-k-test/GCEH-*/bubbleH/0.all-pairs.dat

GCEH all parameters, simulated for BubbleH