public void PerformLocalSearch(Ant <ClusterExampleAssignment> ant)
{
Solution <ClusterExampleAssignment> originalSolution = ant.Solution;
_kmeans.SetAssignment(originalSolution.ToList());
ClusteringSolution clusteringSolution = _kmeans.CreateClusters();
Solution <ClusterExampleAssignment> optimizedSolution = new Solution <ClusterExampleAssignment>();
List <int> optimizedTrail = new List <int>();
foreach (Cluster cluster in clusteringSolution.Clusters)
{
foreach (DataMining.Data.Example example in cluster.Examples)
{
int componentIndex = (example.Index * ClustersNumber) + cluster.Label;
optimizedSolution.Components.Add(new DecisionComponent <ClusterExampleAssignment>(componentIndex, new ClusterExampleAssignment(example.Index, cluster.Label)));
optimizedTrail.Add(componentIndex);
}
}
this.SolutionQualityEvaluator.EvaluateSolutionQuality(optimizedSolution);
if (optimizedSolution.Quality > originalSolution.Quality)
{
ant.Solution = optimizedSolution;
ant.Trail = optimizedTrail;
}
}
private void Initialize(double[][] observations)
{
var algo = new KMeans();
for (var j = 0; j < NumberOfStates; j++)
{
_pi[j] = 1d / NumberOfStates;
}
for (var j = 0; j < NumberOfStates; j++)
{
_tpm[j] = (double[])_pi.Clone();
}
var k = _pi.Length * NumberOfComponents;
var dimentions = observations[0].Length;
algo.CreateClusters(observations, k, KMeans.KMeansDefaultIterations, (k > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
_emission = new Mixture <IMultivariateDistribution> [_pi.Length];
for (int i = 0; i < _pi.Length; i++)
{
_emission[i] = new Mixture <IMultivariateDistribution>(NumberOfComponents, dimentions);
for (int j = 0; j < NumberOfComponents; j++)
{
var mean = algo.ClusterCenters[j + NumberOfComponents * i];
var covariance = algo.ClusterCovariances[j + NumberOfComponents * i];
_emission[i].Components[j] = new NormalDistribution(mean, covariance);
}
}
}
public static ClusteringSolution CreateKMeansClusters(int seed, Dataset dataset, int clustersNumber, ISimilarityMeasure similarityMeasure, int maxIterations, bool fireEvents)
{
DataMining.Utilities.RandomUtility.Initialize(seed);
KMeans kmeans = new KMeans(dataset, clustersNumber, similarityMeasure, maxIterations, fireEvents);
kmeans.Initialize();
kmeans.OnPostIteration += new EventHandler(kmeans_OnPostIteration);
return(kmeans.CreateClusters());
}
static void Main(string[] args)
{
var stopwatch = Stopwatch.StartNew();
KMeans kMeans = new KMeans(40, 4);
kMeans.CreateClusters();
kMeans.PrintClusters();
stopwatch.Stop();
Console.WriteLine($"Elapsed time: {stopwatch.Elapsed.TotalSeconds} seconds.");
Console.Read();
}
public void CreateClusters_FTSEObservations_4Clusters()
{
var util = new TestDataUtils();
var series = util.GetSvcData(util.FTSEFilePath, new DateTime(2010, 12, 18), new DateTime(2011, 12, 18));
var algo = new KMeans();
algo.CreateClusters(series, 4, 200, InitialClusterSelectionMethod.Random);
Assert.AreEqual(4, algo.Clusters.Count);
for (int i = 0; i < algo.Clusters.Count; i++)
{
Assert.IsTrue(algo.Clusters[i].Length > 0);
}
}
private void Initialize(double[][] observations)
{
// Create initial emmissions , TMP and Pi are already created
var algo = new KMeans();
algo.CreateClusters(observations, _pi.Length, KMeans.KMeansDefaultIterations, (_pi.Length > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
_emission = new IMultivariateDistribution[_pi.Length];
for (int i = 0; i < _pi.Length; i++)
{
var mean = algo.ClusterCenters[i];
var covariance = algo.ClusterCovariances[i];
_emission[i] = new NormalDistribution(mean, covariance);
}
}
private void Initialize(double[][] observations)
{
// Create new Emty components
var covariance = new double[_mixture.Dimension, _mixture.Dimension];
var mean = new double[_mixture.Dimension];
var algo = new KMeans();
algo.CreateClusters(observations, _mixture.Components.Length, KMeans.KMeansDefaultIterations, (_mixture.Components.Length > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
for (int i = 0; i < _mixture.Components.Length; i++)
{
mean = algo.ClusterCenters[i];
covariance = algo.ClusterCovariances[i];
_mixture.Components[i] = new NormalDistribution(mean, covariance);
}
}
private NormalDistribution[] CreateEmissions(double[][] observations, int numberOfEmissions)
{
var emissions = new NormalDistribution[numberOfEmissions];
// Create initial emmissions , TMP and Pi are already created
var algo = new KMeans();
algo.CreateClusters(observations, numberOfEmissions, KMeans.KMeansDefaultIterations, (numberOfEmissions > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
for (int i = 0; i < numberOfEmissions; i++)
{
var mean = algo.ClusterCenters[i];
var covariance = algo.ClusterCovariances[i];
emissions[i] = new NormalDistribution(mean, covariance);
}
return(emissions);
}
private void Initialize(double[][] observations)
{
var algo = new KMeans();
var k = _pi.Length * _numberOfComponents;
var dimentions = observations[0].Length;
algo.CreateClusters(observations, k, KMeans.KMeansDefaultIterations, (k > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
_emission = new Mixture <IMultivariateDistribution> [_pi.Length];
for (int i = 0; i < _pi.Length; i++)
{
_emission[i] = new Mixture <IMultivariateDistribution>(_numberOfComponents, dimentions);
for (int j = 0; j < _numberOfComponents; j++)
{
var mean = algo.ClusterCenters[j + _numberOfComponents * i];
var covariance = algo.ClusterCovariances[j + _numberOfComponents * i];
_emission[i].Components[j] = new NormalDistribution(mean, covariance);
Debug.WriteLine("[i,j]=[{0},{1}] Mean Vector {2}", i, j, new Vector(mean));
}
}
}
private Mixture <IMultivariateDistribution>[] CreateEmissions(double[][] observations, int numberOfStates, int numberOfComponents)
{
// Create initial emmissions , TMP and Pi are already created
var algo = new KMeans();
algo.CreateClusters(observations, numberOfStates * numberOfComponents, KMeans.KMeansDefaultIterations, (numberOfStates * numberOfComponents > 3) ? InitialClusterSelectionMethod.Furthest : InitialClusterSelectionMethod.Random);
var emissions = new Mixture <IMultivariateDistribution> [numberOfStates];
for (int i = 0; i < numberOfStates; i++)
{
emissions[i] = new Mixture <IMultivariateDistribution>(numberOfComponents, observations[0].Length);
for (int j = 0; j < numberOfComponents; j++)
{
var mean = algo.ClusterCenters[j + numberOfComponents * i];
var covariance = algo.ClusterCovariances[j + numberOfComponents * i];
emissions[i].Components[j] = new NormalDistribution(mean, covariance);
}
}
return(emissions);
}
