public BiCluster(IEnumerable <long> vector, int id, double distance, BiCluster left, BiCluster right)
{
Vector = vector;
Id = id;
Distance = distance;
Left = left;
Right = right;
}
public BiCluster(IEnumerable<long> vector, int id, double distance, BiCluster left, BiCluster right)
{
Vector = vector;
Id = id;
Distance = distance;
Left = left;
Right = right;
}
public static BiCluster HCluster(IEnumerable<WordVector> words)
{
var wordsList = words.ToArray();
var distances = new Dictionary<Pair, double>();
int currentClustId = -1;
List<BiCluster> clust = new List<BiCluster>(wordsList.Length);
for (int i = 0; i < wordsList.Length; i++)
clust.Add(new BiCluster(wordsList[i].Entities.Values, i, 0.0));
while (clust.Count > 1)
{
Pair lowestpair = new Pair(0, 1);
var closest = Pearson(clust[0].Vector, clust[1].Vector);
// loop through every pair looking for the smallest distance
for (int i = 0; i < clust.Count; i++)
{
for (int j = i + 1; j < clust.Count; j++)
{
// distances is the cache of distance calculations
if (!distances.ContainsKey(new Pair(clust[i].Id, clust[j].Id)))
{
// Calculate the pearson value
var val = Pearson(clust[i].Vector, clust[j].Vector);
distances.Add(new Pair(clust[i].Id, clust[j].Id), val);
}
double d = distances[new Pair(clust[i].Id, clust[j].Id)];
if (d < closest)
{
closest = d;
lowestpair = new Pair(i, j);
}
}
}
// calculate the average of the two clusters
List<long> mergevec = new List<long>();
long[] vectorArray = clust[0].Vector.ToArray();
for (int i = 0; i < vectorArray.Length; i++)
{
long[] leftVectorArray = clust[lowestpair.Left].Vector.ToArray();
long[] rightVectorArray = clust[lowestpair.Right].Vector.ToArray();
mergevec.Add((leftVectorArray[i] + rightVectorArray[i]) / 2);
}
//create the new cluster
BiCluster newCluster = new BiCluster(mergevec, currentClustId, closest, clust[lowestpair.Left], clust[lowestpair.Right]);
//cluster ids that weren't in the original set are negative
currentClustId--;
clust.Remove(clust[lowestpair.Left]);
clust.Remove(clust[lowestpair.Right]);
clust.Add(newCluster);
}
return clust[0];
}
public static BiCluster HCluster(IEnumerable <WordVector> words)
{
var wordsList = words.ToArray();
var distances = new Dictionary <Pair, double>();
int currentClustId = -1;
List <BiCluster> clust = new List <BiCluster>(wordsList.Length);
for (int i = 0; i < wordsList.Length; i++)
{
clust.Add(new BiCluster(wordsList[i].Entities.Values, i, 0.0));
}
while (clust.Count > 1)
{
Pair lowestpair = new Pair(0, 1);
var closest = Pearson(clust[0].Vector, clust[1].Vector);
// loop through every pair looking for the smallest distance
for (int i = 0; i < clust.Count; i++)
{
for (int j = i + 1; j < clust.Count; j++)
{
// distances is the cache of distance calculations
if (!distances.ContainsKey(new Pair(clust[i].Id, clust[j].Id)))
{
// Calculate the pearson value
var val = Pearson(clust[i].Vector, clust[j].Vector);
distances.Add(new Pair(clust[i].Id, clust[j].Id), val);
}
double d = distances[new Pair(clust[i].Id, clust[j].Id)];
if (d < closest)
{
closest = d;
lowestpair = new Pair(i, j);
}
}
}
// calculate the average of the two clusters
List <long> mergevec = new List <long>();
long[] vectorArray = clust[0].Vector.ToArray();
for (int i = 0; i < vectorArray.Length; i++)
{
long[] leftVectorArray = clust[lowestpair.Left].Vector.ToArray();
long[] rightVectorArray = clust[lowestpair.Right].Vector.ToArray();
mergevec.Add((leftVectorArray[i] + rightVectorArray[i]) / 2);
}
//create the new cluster
BiCluster newCluster = new BiCluster(mergevec, currentClustId, closest, clust[lowestpair.Left], clust[lowestpair.Right]);
//cluster ids that weren't in the original set are negative
currentClustId--;
clust.Remove(clust[lowestpair.Left]);
clust.Remove(clust[lowestpair.Right]);
clust.Add(newCluster);
}
return(clust[0]);
}
