public static List <DataVector> AggregateTrueMeansToFileSystem(int partitionsNum, int clustersNum, string executionDirectory)
{
List <PartialMean> partialMeans = new List <PartialMean>();
for (int i = 0; i < partitionsNum; i++)
{
// should be replaced with Group Communication
string path = Path.Combine(executionDirectory, Constants.DataDirectory, Constants.PartialMeanFilePrefix + i.ToString(CultureInfo.InvariantCulture));
FileStream file = new FileStream(path, FileMode.Open, FileAccess.Read, FileShare.Read);
using (StreamReader reader = new StreamReader(file))
{
int index = 0;
while (!reader.EndOfStream)
{
string line = reader.ReadLine();
if (index++ < clustersNum)
{
partialMeans.Add(PartialMean.FromString(line));
}
}
reader.Close();
}
}
List <DataVector> newCentroids = new List <DataVector>();
for (int i = 0; i < clustersNum; i++)
{
List <PartialMean> means = partialMeans.Where(m => m.Mean.Label == i).ToList();
newCentroids.Add(PartialMean.AggreatedMean(means));
}
return(newCentroids);
}
public ProcessedResults Reduce(IEnumerable <ProcessedResults> elements)
{
List <PartialMean> aggregatedMeans = new List <PartialMean>();
List <PartialMean> totalList = new List <PartialMean>();
float aggregatedLoss = 0;
foreach (var element in elements)
{
totalList.AddRange(element.Means);
aggregatedLoss += element.Loss;
}
// we infer the value of K from the labeled data
int clustersNum = totalList.Max(p => p.Mean.Label) + 1;
for (int i = 0; i < clustersNum; i++)
{
List <PartialMean> means = totalList.Where(m => m.Mean.Label == i).ToList();
aggregatedMeans.Add(new PartialMean(PartialMean.AggreatedMean(means), means.Count));
}
ProcessedResults returnMeans = new ProcessedResults(aggregatedMeans, aggregatedLoss);
Logger.Log(Level.Info, "The true means aggregated by the reduce function: " + returnMeans);
return(returnMeans);
}
public static DataVector AggreatedMean(List <PartialMean> means)
{
if (means == null || means.Count == 0)
{
throw new ArgumentException("means");
}
PartialMean mean = means[0];
for (int i = 1; i < means.Count; i++)
{
mean = mean.CombinePartialMean(means[i]);
}
return(mean.Mean);
}
private PartialMean CombinePartialMean(PartialMean other)
{
PartialMean aggreatedMean = new PartialMean();
if (other == null)
{
throw new ArgumentNullException("other");
}
if (Mean.Label != other.Mean.Label)
{
throw new ArgumentException("cannot combine partial means with different labels");
}
aggreatedMean.Size = Size + other.Size;
aggreatedMean.Mean = Mean.MultiplyScalar(Size).Add(other.Mean.MultiplyScalar(other.Size)).Normalize(aggreatedMean.Size);
return(aggreatedMean);
}
public List <PartialMean> ComputePartialMeans()
{
List <PartialMean> partialMeans = new PartialMean[_clustersNum].ToList();
for (int i = 0; i < _clustersNum; i++)
{
List <DataVector> slices = _dataPartition.DataVectors.Where(d => d.Label == i).ToList();
DataVector average = new DataVector(_dataPartition.DataVectors[0].Dimension);
if (slices.Count > 1)
{
average = DataVector.Mean(slices);
}
average.Label = i;
partialMeans[i] = new PartialMean(average, slices.Count);
}
return(partialMeans);
}
private List <PartialMean> ComputePartialMeans()
{
Logger.Log(Level.Verbose, "cluster number " + _clustersNum);
List <PartialMean> partialMeans = new PartialMean[_clustersNum].ToList();
for (int i = 0; i < _clustersNum; i++)
{
List <DataVector> slices = _dataPartition.DataVectors.Where(d => d.Label == i).ToList();
DataVector average = new DataVector(_dataPartition.DataVectors[0].Dimension);
if (slices.Count > 1)
{
average = DataVector.Mean(slices);
}
average.Label = i;
partialMeans[i] = new PartialMean(average, slices.Count);
Logger.Log(Level.Info, "Adding to partial means list: " + partialMeans[i]);
}
return(partialMeans);
}
private List<PartialMean> ComputePartialMeans()
{
Logger.Log(Level.Verbose, "cluster number " + _clustersNum);
List<PartialMean> partialMeans = new PartialMean[_clustersNum].ToList();
for (int i = 0; i < _clustersNum; i++)
{
List<DataVector> slices = _dataPartition.DataVectors.Where(d => d.Label == i).ToList();
DataVector average = new DataVector(_dataPartition.DataVectors[0].Dimension);
if (slices.Count > 1)
{
average = DataVector.Mean(slices);
}
average.Label = i;
partialMeans[i] = new PartialMean(average, slices.Count);
Logger.Log(Level.Info, "Adding to partial means list: " + partialMeans[i]);
}
return partialMeans;
}
private PartialMean CombinePartialMean(PartialMean other)
{
PartialMean aggreatedMean = new PartialMean();
if (other == null)
{
throw new ArgumentNullException("other");
}
if (Mean.Label != other.Mean.Label)
{
throw new ArgumentException("cannot combine partial means with different labels");
}
aggreatedMean.Size = Size + other.Size;
aggreatedMean.Mean = Mean.MultiplyScalar(Size).Add(other.Mean.MultiplyScalar(other.Size)).Normalize(aggreatedMean.Size);
return aggreatedMean;
}
