private void RunKMeans(string name, string dirName, string alignFile = null, DCDFile dcd = null)
{
DateTime cpuPart1 = DateTime.Now;
ClusterOutput clustOut;
DistanceMeasure distance = null;
if (dcd == null)
{
distance = CreateMeasure(name, dirName, opt.kmeans.kDistance, opt.kmeans.kAtoms, opt.kmeans.reference1Djury,
alignFile, opt.kmeans.hammingProfile, opt.kmeans.jury1DProfile);
}
else
{
distance = CreateMeasureForDCD(dcd, opt.kmeans.kDistance, opt.kmeans.kAtoms, opt.kmeans.reference1Djury,
opt.kmeans.alignmentFileName, opt.kmeans.hammingProfile, opt.kmeans.jury1DProfile);
}
kMeans km;
km = new kMeans(distance, opt.kmeans.kMeans_init);
if (beginJob != null)
{
beginJob(currentProcessName, km.ToString(), dirName, distance.ToString());
}
progressDic.Add(name, km);
distance.InitMeasure();
DateTime cpuPart2 = DateTime.Now;
clType = km.ToString();
if ((int)opt.kmeans.maxK <= 1)
{
throw new Exception("k in k-Means must be bigger then 1, right now is: " + (int)opt.kmeans.maxK);
}
if (distance.structNames.Count < 10)
{
throw new Exception("Number of structures to cluster must be bigger then 10 right now is: " + distance.structNames.Count);
}
clustOut = km.kMeansLevel((int)opt.kmeans.maxK, opt.kmeans.maxIter, new List <string>(distance.structNames.Keys));
UpdateOutput(name, dirName, alignFile, clustOut, distance.ToString(), cpuPart1, cpuPart2, km);
GC.SuppressFinalize(distance);
}
private void RunHKMeans(string name, string dirName, string alignFile = null, DCDFile dcd = null)
{
DateTime cpuPart1 = DateTime.Now;
ClusterOutput clustOut = null;
DistanceMeasure distance = null;
if (dcd == null)
{
distance = CreateMeasure(name, dirName, opt.hierarchical.distance, opt.hierarchical.atoms, opt.hierarchical.reference1DjuryKmeans,
alignFile, opt.hierarchical.hammingProfile, opt.hierarchical.jury1DProfileKmeans);
}
else
{
distance = CreateMeasureForDCD(dcd, opt.hierarchical.distance, opt.hierarchical.atoms, opt.hierarchical.reference1DjuryKmeans,
opt.hierarchical.alignmentFileName, opt.hierarchical.hammingProfile, opt.hierarchical.jury1DProfileKmeans);
}
kMeans km;
km = new kMeans(distance, true);
if (beginJob != null)
{
beginJob(currentProcessName, km.ToString(), dirName, distance.ToString());
}
progressDic.Add(name, km);
DateTime cpuPart2 = DateTime.Now;
distance.InitMeasure();
clType = km.ToString();
km.BMIndex = opt.hierarchical.indexDB;
km.threshold = opt.hierarchical.numberOfStruct;
km.maxRepeat = opt.hierarchical.repeatTime;
km.maxK = opt.hierarchical.maxK;
clustOut = km.HierarchicalKMeans();
UpdateOutput(name, dirName, alignFile, clustOut, distance.ToString(), cpuPart1, cpuPart2, km);
}
ClusterOutput DivideSpace1DJury(List <string> list)
{
ClusterOutput outC, finalOut = new ClusterOutput();
outC = jury.JuryOptWeights(list);
double dev = CalcStandDev(outC);
List <string> clust1 = new List <string>();
List <string> clust2 = new List <string>();
dev += dev;
if (useKMeans)
{
clust1.Add(outC.juryLike[0].Key);
clust1.Add(outC.juryLike[outC.juryLike.Count - 1].Key);
kMeans km = new kMeans(dMeasure);
return(km.kMeansRun(kMeansIter, list, clust1));
}
else
{
for (int i = 0; i < outC.juryLike.Count; i++)
{
if (i < outC.juryLike.Count / 2)
{
clust1.Add(outC.juryLike[i].Key);
}
else
{
clust2.Add(outC.juryLike[i].Key);
}
}
finalOut.clusters = new List <List <string> >();
finalOut.clusters.Add(clust1);
finalOut.clusters.Add(clust2);
}
return(finalOut);
}
private void RunKMeans(string name, string dirName, string alignFile=null, DCDFile dcd = null)
{
DateTime cpuPart1 = DateTime.Now;
ClusterOutput clustOut;
DistanceMeasure distance = null;
if(dcd==null)
distance = CreateMeasure(name,dirName,opt.kmeans.kDistance, opt.kmeans.kAtoms, opt.kmeans.reference1Djury,
alignFile, opt.kmeans.hammingProfile, opt.kmeans.jury1DProfile);
else
distance =CreateMeasureForDCD(dcd, opt.kmeans.kDistance, opt.kmeans.kAtoms, opt.kmeans.reference1Djury,
opt.kmeans.alignmentFileName, opt.kmeans.hammingProfile, opt.kmeans.jury1DProfile);
kMeans km;
km = new kMeans(distance, opt.kmeans.kMeans_init);
progressDic.Add(name, km);
distance.InitMeasure();
DateTime cpuPart2 = DateTime.Now;
clType = km.ToString();
if ((int)opt.kmeans.maxK <= 1)
throw new Exception("k in k-Means must be bigger then 1, right now is: " + (int)opt.kmeans.maxK);
if (distance.structNames.Count < 10)
throw new Exception("Number of structures to cluster must be bigger then 10 right now is: " + distance.structNames.Count);
clustOut = km.kMeansLevel((int)opt.kmeans.maxK, opt.kmeans.maxIter,new List <string>(distance.structNames.Keys));
UpdateOutput(name, dirName,alignFile, clustOut, distance.ToString(), cpuPart1, cpuPart2,km);
GC.SuppressFinalize(distance);
}
private void RunHKMeans(string name, string dirName, string alignFile=null,DCDFile dcd = null)
{
DateTime cpuPart1 = DateTime.Now;
ClusterOutput clustOut = null;
DistanceMeasure distance = null;
if(dcd==null)
distance = CreateMeasure(name,dirName,opt.hierarchical.distance, opt.hierarchical.atoms, opt.hierarchical.reference1DjuryKmeans,
alignFile, opt.hierarchical.hammingProfile, opt.hierarchical.jury1DProfileKmeans);
else
distance = CreateMeasureForDCD(dcd, opt.hierarchical.distance, opt.hierarchical.atoms, opt.hierarchical.reference1DjuryKmeans,
opt.hierarchical.alignmentFileName, opt.hierarchical.hammingProfile, opt.hierarchical.jury1DProfileKmeans);
kMeans km;
km = new kMeans(distance,true);
progressDic.Add(name, km);
DateTime cpuPart2 = DateTime.Now;
distance.InitMeasure();
clType = km.ToString();
km.BMIndex = opt.hierarchical.indexDB;
km.threshold = opt.hierarchical.numberOfStruct;
km.maxRepeat = opt.hierarchical.repeatTime;
km.maxK = opt.hierarchical.maxK;
clustOut = km.HierarchicalKMeans();
UpdateOutput(name, dirName,alignFile,clustOut, distance.ToString(), cpuPart1, cpuPart2, km);
}
ClusterOutput DivideSpaceHamming(List <string> list)
{
ClusterOutput output = new ClusterOutput();
Dictionary <string, int> aux = new Dictionary <string, int>();
ClusterOutput outC = jury.JuryOptWeights(list);
if (outC == null)
{
return(null);
}
List <string> clust1 = new List <string>();
List <string> clust2 = new List <string>();
//consensus.ToConsensusStates(list, jury.GetStructureStates(outC.juryLike[0].Key));
consensus.ToConsensusStates(list, outC.juryLike[0].Key);
foreach (var item in list)
{
aux.Add(item, consensus.distanceOrdered[item]);
}
if (useKMeans)
{
clust1.Add(outC.juryLike[0].Key);
var item = aux.OrderByDescending(j => j.Value);
int dist = item.First().Value;
foreach (var ll in item)
{
if (dist != ll.Value)
{
break;
}
else
{
clust2.Add(ll.Key);
}
}
Dictionary <string, double> dic = new Dictionary <string, double>();
foreach (var it in outC.juryLike)
{
dic.Add(it.Key, it.Value);
}
double min = Double.MaxValue;
string rem = "";
foreach (var it in clust2)
{
if (dic[it] < min)
{
min = dic[it];
rem = it;
}
}
clust1.Add(rem);
kMeans km = new kMeans(dMeasure);
return(km.kMeansRun(kMeansIter, list, clust1));
}
else
{
int i = 0;
foreach (var item in aux.OrderBy(j => j.Value))
{
if (i < list.Count() / 2)
{
clust1.Add(item.Key);
}
else
{
clust2.Add(item.Key);
}
i++;
}
output.clusters = new List <List <string> >();
output.clusters.Add(clust1);
output.clusters.Add(clust2);
}
return(output);
}
ClusterOutput DivideSpaceKmeans(List <string> list)
{
kMeans km = new kMeans(dMeasure);
return(km.kMeansLevel(2, 30, list));
}
ClusterOutput DivideSpaceHamming(List<string> list)
{
ClusterOutput output=new ClusterOutput();
Dictionary<string, int> aux = new Dictionary<string, int>();
ClusterOutput outC = jury.JuryOptWeights(list);
if (outC == null)
return null;
List<string> clust1 = new List<string>();
List<string> clust2 = new List<string>();
//consensus.ToConsensusStates(list, jury.GetStructureStates(outC.juryLike[0].Key));
consensus.ToConsensusStates(list,outC.juryLike[0].Key);
foreach (var item in list)
aux.Add(item, consensus.distanceOrdered[item]);
if (useKMeans)
{
clust1.Add(outC.juryLike[0].Key);
var item = aux.OrderByDescending(j => j.Value);
int dist = item.First().Value;
foreach (var ll in item)
{
if (dist != ll.Value)
break;
else
clust2.Add(ll.Key);
}
Dictionary<string, double> dic = new Dictionary<string, double>();
foreach (var it in outC.juryLike)
dic.Add(it.Key, it.Value);
double min = Double.MaxValue;
string rem = "";
foreach (var it in clust2)
{
if (dic[it] < min)
{
min = dic[it];
rem = it;
}
}
clust1.Add(rem);
kMeans km = new kMeans(dMeasure);
return km.kMeansRun(kMeansIter,list, clust1);
}
else
{
int i = 0;
foreach (var item in aux.OrderBy(j => j.Value))
{
if (i < list.Count() / 2)
clust1.Add(item.Key);
else
clust2.Add(item.Key);
i++;
}
output.clusters = new List<List<string>>();
output.clusters.Add(clust1);
output.clusters.Add(clust2);
}
return output;
}
ClusterOutput DivideSpaceKmeans(List<string> list)
{
kMeans km = new kMeans(dMeasure);
return km.kMeansLevel(2,30, list);
}
ClusterOutput DivideSpace1DJury(List<string> list)
{
ClusterOutput outC,finalOut=new ClusterOutput();
outC=jury.JuryOptWeights(list);
double dev = CalcStandDev(outC);
List<string> clust1 = new List<string>();
List<string> clust2 = new List<string>();
dev += dev;
if (useKMeans)
{
clust1.Add(outC.juryLike[0].Key);
clust1.Add(outC.juryLike[outC.juryLike.Count - 1].Key);
kMeans km = new kMeans(dMeasure);
return km.kMeansRun(kMeansIter,list, clust1);
}
else
{
for (int i = 0; i < outC.juryLike.Count; i++)
if(i<outC.juryLike.Count / 2)
clust1.Add(outC.juryLike[i].Key);
else
clust2.Add(outC.juryLike[i].Key);
finalOut.clusters = new List<List<string>>();
finalOut.clusters.Add(clust1);
finalOut.clusters.Add(clust2);
}
return finalOut;
}
