private static void RunDatabaseAccessesEntropy(Random random)
{
int numOfNodes = 10;
var window = 14;
var vectorLength = 11654;
var distrubteUsers = UsersDistributing.RoundRobin();
var approximation = new MultiplicativeApproximation(0.02);
EntropyRunner.RunDatabaseAccesses(random, numOfNodes, window, approximation, vectorLength, distrubteUsers, databaseAccessesPath, resultDir);
}
public static void RunDatabaseAccesses(Random rnd, int numOfNodes,
int window,
ApproximationType approximation, int vectorLength,
UsersDistributing distributing,
string databaseAccessesPath, string resultDir)
{
var resultPath =
PathBuilder.Create(resultDir, "Entropy")
.AddProperty("Dataset", "DatabaseAccesses")
.AddProperty("Nodes", numOfNodes.ToString())
.AddProperty("Window", window.ToString())
.AddProperty("Approximation", approximation.AsString())
.AddProperty("DistributingMethod", distributing.Name)
.ToPath("csv");
var hashUser = new Func <int, int>(userId => userId % numOfNodes);
var entropy = new EntropyFunction(vectorLength);
using (var resultCsvFile = AutoFlushedTextFile.Create(resultPath, AccumaltedResult.Header(numOfNodes)))
using (var databaseAccessesStatistics = DatabaseAccessesStatistics.Init(databaseAccessesPath, numOfNodes, window, distributing.DistributeFunc))
{
var initProbabilityVectors = databaseAccessesStatistics.InitProbabilityVectors();
if (!initProbabilityVectors.All(v => v.Sum().AlmostEqual(1.0, 0.000001)))
{
throw new Exception();
}
var multiRunner = MultiRunner.InitAll(initProbabilityVectors, numOfNodes, vectorLength,
approximation, entropy.MonitoredFunction);
while (databaseAccessesStatistics.TakeStep())
{
var changeProbabilityVectors = databaseAccessesStatistics.GetChangeProbabilityVectors();
if (!changeProbabilityVectors.All(v => v.Sum().AlmostEqual(0.0, 0.000001)))
{
throw new Exception();
}
multiRunner.Run(changeProbabilityVectors, rnd, true)
.Select(r => r.AsCsvString())
.ForEach(resultCsvFile.WriteLine);
}
}
}
public static void RunDatabaseAccesses(Random rnd, int numOfNodes, int window,
ApproximationType approximation, int width,
int height, UsersDistributing distributing,
string databaseAccessesPath, string resultDir)
{
var vectorLength = width * height;
var hashFunction = FourwiseIndepandantFunction.Init(rnd);
var hashFunctionsTable = HashFunctionTable.Init(numOfNodes, vectorLength, hashFunction);
var secondMomentFunction = new SecondMoment(width, height);
var resultPath =
PathBuilder.Create(resultDir, "AMS_F2")
.AddProperty("Dataset", "DatabaseAccesses")
.AddProperty("Nodes", numOfNodes.ToString())
.AddProperty("VectorLength", vectorLength.ToString())
.AddProperty("Width", width.ToString())
.AddProperty("Height", height.ToString())
.AddProperty("Window", window.ToString())
.AddProperty("Distributing", distributing.Name)
.AddProperty("Approximation", approximation.AsString())
.ToPath("csv");
using (var resultCsvFile = AutoFlushedTextFile.Create(resultPath, AccumaltedResult.Header(numOfNodes)))
using (var databaseAccessesStatistics = DatabaseAccessesStatistics.Init(databaseAccessesPath, numOfNodes, window, distributing.DistributeFunc))
{
var initCountVectors = databaseAccessesStatistics.InitCountVectors();
var initVectors = hashFunction.TransformToAMSSketch(initCountVectors, vectorLength, hashFunctionsTable);
var multiRunner = MultiRunner.InitAll(initVectors, numOfNodes, vectorLength,
approximation, secondMomentFunction.MonitoredFunction);
while (databaseAccessesStatistics.TakeStep())
{
var changeCountVectors = databaseAccessesStatistics.GetChangeCountVectors();
var changeVectors = hashFunction.TransformToAMSSketch(changeCountVectors, vectorLength, hashFunctionsTable);
multiRunner.Run(changeVectors, rnd, true)
.Select(r => r.AsCsvString())
.ForEach(resultCsvFile.WriteLine);
}
}
}
