/// <summary>
/// Run all analytics experiments.
/// </summary>
public static void AllAnalyticsExperiments(string filter = "*")
{
string path = Input("Please enter the path to folder with graph files", string.Copy);
string[] filePaths = Directory.GetFiles(path, filter + ".xml");
string outPath = Path.GetDirectoryName(path) + @"\Analytics";
if (!Directory.Exists(outPath))
{
Directory.CreateDirectory(outPath);
}
foreach (var filePath in filePaths)
{
var graph = GraphLoader.LoadGraphML(filePath, int.Parse, int.Parse);
var exp1 = Experiments.DistanceProbabilityMassFunction(graph);
Experiment.SaveSVG(outPath + @"\" + string.Join("_", exp1.Meta) + ".svg", exp1.Plot(0, double.NaN));
exp1.SaveTSV(outPath + @"\" + string.Join("_", exp1.Meta) + ".tsv");
var exp2 = Experiments.BisimulationPartitionSize(graph);
Experiment.SaveSVG(outPath + @"\" + string.Join("_", exp2.Meta) + ".svg", exp2.Plot(0, double.NaN));
exp2.SaveTSV(outPath + @"\" + string.Join("_", exp2.Meta) + ".tsv");
var exp3 = Experiments.PartitionBlockDistribution(graph);
Experiment.SaveSVG(outPath + @"\" + string.Join("_", exp3.Meta) + ".svg", exp3.Plot(0, double.NaN));
exp3.SaveTSV(outPath + @"\" + string.Join("_", exp3.Meta) + ".tsv");
}
}
/// <summary>
/// Run all performance experiments sequentially.
/// </summary>
public static void AllPerformanceExperiments(string filter = "*")
{
string path = Input("Please enter the path to folder with graph files", string.Copy);
int M = Input("Maximum number of machines?", int.Parse);
int algorithm = Select("Algorithm?", new string[] { "Exact (explore)", "Estimate (explore)", "Exact (random)", "Estimate (random)", "All of the above" });
string outPath = Path.GetDirectoryName(path) + @"\Performance";
if (!Directory.Exists(outPath))
{
Directory.CreateDirectory(outPath);
}
string[] filePaths = Directory.GetFiles(path, filter + ".xml");
foreach (var filePath in filePaths)
{
var graph = GraphLoader.LoadGraphML(filePath, int.Parse, int.Parse);
Experiment[] experiments = null;
switch (algorithm)
{
case 0:
experiments = Experiments.MeasureDistributedPerformanceExact(graph, M, DistributedUtils.ExploreSplit, "Explore");
break;
case 1:
experiments = Experiments.MeasureDistributedPerformanceEstimate(graph, M, DistributedUtils.ExploreSplit, "Explore");
break;
case 2:
experiments = Experiments.MeasureDistributedPerformanceExact(graph, M, DistributedUtils.RandomSplit, "Random");
break;
case 3:
experiments = Experiments.MeasureDistributedPerformanceEstimate(graph, M, DistributedUtils.RandomSplit, "Random");
break;
case 4:
var exps1 = Experiments.MeasureDistributedPerformanceExact(graph, M, DistributedUtils.ExploreSplit, "Explore");
var exps2 = Experiments.MeasureDistributedPerformanceEstimate(graph, M, DistributedUtils.ExploreSplit, "Explore");
var exps3 = Experiments.MeasureDistributedPerformanceExact(graph, M, DistributedUtils.RandomSplit, "Random");
var exps4 = Experiments.MeasureDistributedPerformanceEstimate(graph, M, DistributedUtils.RandomSplit, "Random");
experiments = exps1.Concat(exps2).Concat(exps3).Concat(exps4).ToArray();
break;
}
foreach (var experiment in experiments)
{
var plot = experiment.Plot(0, double.NaN);
experiment.GetHorizontalAxis(plot).MinorTickSize = 0;
experiment.GetHorizontalAxis(plot).MajorStep = 1;
Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", plot);
experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
}
}
}
public static void Bla()
{
// Merge graphs in a folder with a single source node and sink node
string path = Program.Input("Please enter the path to folder with graph files", string.Copy);
var outPath = Program.Input("Out path?", string.Copy);
string[] filePaths = Directory.GetFiles(path, "*.xml");
var finalGraph = new MultiDirectedGraph <int, int>();
var finalSources = new List <int>();
var finalSinks = new List <int>();
int count = 0;
foreach (var filePath in filePaths)
{
var graph = GraphLoader.LoadGraphML(filePath, int.Parse, int.Parse);
foreach (var node in graph.Nodes)
{
finalGraph.AddNode(node + count, graph.NodeLabel(node));
}
foreach (var edge in graph.Edges)
{
var s = graph.Source(edge);
var t = graph.Target(edge);
var l = graph.EdgeLabel(edge);
finalGraph.AddEdge(s + count, t + count, l);
}
var sources = graph.Nodes.Where(u => graph.In(u).Count() == 0);
var sinks = graph.Nodes.Where(u => graph.Out(u).Count() == 0);
if (sources.Count() != 1 || sinks.Count() != 1)
{
throw new Exception();
}
finalSources.Add(sources.First() + count);
finalSinks.Add(sinks.First() + count);
count += graph.NumNodes;
}
finalGraph.MergeNodes(finalSources);
finalGraph.MergeNodes(finalSinks);
GraphConverter.SaveToGraphML(finalGraph, outPath);
}
/// <summary>
/// Entry point.
/// </summary>
/// <param name="args"></param>
public static void Main(string[] args)
{
//*
Dummy.Fix();
return;
//*/
/*
* AllPerformanceExperiments();
* return;
* //*/
//*
AllAnalyticsExperiments("Petrinet_no_labels");
return;
//*/
// Ask for graph file
string path = Input("Please enter the path to the graph file", string.Copy);
string outPath = Path.GetDirectoryName(path) + @"\..\Results";
// Load graph and labels
var graph = GraphLoader.LoadGraphML(path, int.Parse, int.Parse);
/*
* var partitioner = new GraphPartitioner<int, int>(graph);
* var distributedPartitioner = new DistributedGraphPartitioner<int, int>(1, graph);
* //*/
/*
* var estim = GraphGenerator.ReducedGraph(graph, distributedPartitioner.ExactBisimulationReduction);
* var exact = GraphGenerator.ReducedGraph(graph, partitioner.EstimateBisimulationReduction);
* //*/
//*
// GraphConverter.SaveToGraphML(estim, Path.GetDirectoryName(path) + @"\" + graph.Name + "_estim.xml");
// GraphConverter.SaveToGraphML(exact, Path.GetDirectoryName(path) + @"\" + graph.Name + "_exact.xml");
// GraphConverter.SaveToGraphML(coarse, Path.GetDirectoryName(path) + @"\" + graph.Name + "_coarse.xml");
//*/
// Samplers
var samplers = new Func <double, MultiDirectedGraph <int, int> >[]
{
//* Normal samplers
p => graph.Induce(graph.RN((int)(p * graph.NumNodes))),
p => graph.Induce(graph.RE((int)(p * graph.NumEdges))),
// p => graph.Induce(graph.LowDegreeFirst((int)(p * graph.NumNodes))),
// p => graph.Induce(graph.GreedyLabels((int)(p * graph.NumNodes))),
p => graph.Induce(graph.DistinctLabelsSB((int)(p * graph.NumNodes))),
p => graph.Induce(graph.QueuedSampler <int, int, FifoQueue <int> >((int)(p * graph.NumNodes))),
// p => graph.Induce(graph.QueuedSampler<int, int, LifoQueue<int>>((int)(p * graph.NumNodes))),
// p => graph.Induce(graph.QueuedSampler<int, int, AiroQueue<int>>((int)(p * graph.NumNodes))),
p => graph.Induce(graph.RandomWalkTeleport((int)(p * graph.NumNodes), 0.1)),
//*/
/* Approximation samplers
* p => estim.Induce(estim.RN((int)(p * graph.NumNodes))),
* p => estim.Induce(estim.RE((int)(p * graph.NumEdges))),
* // p => estim.Induce(estim.LowDegreeFirst((int)(p * graph.NumNodes))),
* // p => estim.Induce(estim.GreedyLabels((int)(p * graph.NumNodes))),
* p => estim.Induce(estim.DistinctLabelsSB((int)(p * graph.NumNodes))),
* p => estim.Induce(estim.QueuedSampler<int, int, FifoQueue<int>>((int)(p * graph.NumNodes))),
* // p => estim.Induce(estim.QueuedSampler<int, int, LifoQueue<int>>((int)(p * graph.NumNodes))),
* // p => estim.Induce(estim.QueuedSampler<int, int, AiroQueue<int>>((int)(p * graph.NumNodes))),
* p => estim.Induce(estim.RandomWalkTeleport((int)(p * graph.NumNodes), 0.1)),
* //*/
};
// Sampler names
var samplerNames = new string[]
{
"RN",
"RE",
// "LDF",
// "GL",
"DLSB",
"BFS",
// "DFS",
// "RFS",
"RWT",
};
//*/
//* Run many bisimulation experiments in batch
var partitioner = new GraphPartitioner <int, int>(graph);
var k_max = partitioner.MultilevelExactBisimulationReduction().Count - 1;
for (int k = 0; k <= k_max; k++)
{
for (int i = 0; i < samplers.Length; i++)
{
var sampler = samplers[i];
var samplerName = samplerNames[i];
// var experiment = Experiments.StandardBisimulationMetrics(graph, samplerName, sampler, k);
// Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", experiment.Plot(0.0, 1.0));
// experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
var experiment = Experiments.WeightedBisimulationMetrics(graph, samplerName, sampler, k);
Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", experiment.Plot(0.0, 1.0));
experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
Console.WriteLine("[" + DateTime.Now.ToString("HH:mm:ss") + "] k=" + k + " sampler=" + samplerName);
}
;
}
//*/
/*
* // var experiment = Analytics.ReachabilityProbabilityMassFunction(graph);
* // var experiment = Experiments.BisimulationPartitionBlockCounts(graph, labels);
* // var experiment = Experiments.FindKMax(graph, labels);
* // var experiment = Experiments.TreeGeneratorPartitionSize(1, 20);
* PlotForm plotForm = new PlotForm();
* plotForm.Display(experiment.Plot(double.NaN, double.NaN));
* plotForm.ShowDialog();
* //*/
/* Get k
* int k = Input("Please enter a value for k in k-bisimulation", int.Parse);
* //*/
/*
* var experiment = Experiments.DistanceProbabilityMassFunction(graph);
* Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", experiment.Plot(0, double.NaN));
* experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
* //*/
/* Run bisimulation experiment
* var experiment = Experiments.StandardBisimulationMetrics(graph, labels, () => GraphSampler.Funny(graph, labels));
* Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", experiment.Plot(0.0, 1.0));
* experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
* //*/
/*
* var experiment = Experiments.BisimulationPartitionSize(graph, labels);
* // var experiment = Experiments.PartitionBlockDistribution(graph, labels);
* Experiment.SaveSVG(outPath + @"\" + string.Join("_", experiment.Meta) + ".svg", experiment.Plot(double.NaN, double.NaN));
* experiment.SaveTSV(outPath + @"\" + string.Join("_", experiment.Meta) + ".tsv");
* //*/
/*
* PlotForm plotForm = new PlotForm();
* plotForm.Display(experiment.Plot(0, double.NaN));
* plotForm.ShowDialog();
* //*/
}
