//this version of loader is for cluster distribution
//take in the number of nodes, and current node number
//we add in data only if it corresponds to our current node number
public TsvFileLoader(RankingTSVFile<MsnData> tsvFile, Random r, double rangeLower, double rangeUpper)
{
DataNullProc<MsnData> dataNullPro = new DataNullProc<MsnData>();
IDataEnum<MsnData, MsnData, DataNullProc<MsnData>> tsvDataEnum = new TsvDataStream<MsnData, MsnData, DataNullProc<MsnData>>(tsvFile, dataNullPro);
int gId = 0;
featureDataMatrix = new DataMatrixArray<float>();
List<float> labelList = new List<float>();
List<int> groupIdList = new List<int>();
//for each query in the set of queries
foreach (MsnData d in tsvDataEnum)
{
//if it falls in the "Test set" range, load it into the test set
double rDouble = r.NextDouble();
if (rDouble >= rangeLower && rDouble < rangeUpper)
{
//load the example
if (featureNames == null)
{
featureNames = new string[d.Feature.NumColumns];
for (int i = 0; i < d.Feature.NumColumns; i++)
{
featureNames[i] = tsvFile.ColumnNames[d.Feature.Parser.columnIndex(i)];
}
}
//get labels
for (int i = 0; i < d.Labels.Data.NumRows; i++)
{
labelList.Add(d.Labels.Data.GetValue(i, 0));
groupIdList.Add(gId); // group index/Id
}
//get feature data
featureDataMatrix.Add(d.Feature.Data);
gId++;
}
//otherwise, skip that query
}
int numRows = labelList.Count;
labels = new float[numRows];
groupId = new int[numRows];
for (int i = 0; i < numRows; i++)
{
labels[i] = labelList[i];
groupId[i] = groupIdList[i];
}
}
public static RankingTSVFile<MsnData> CreateTsvFile(string tsvFileName,
IParser<string> metaParser, IParser<float> labelParser, IParser<float> featureParser,
IGroupBoundary groupBoundary)
{
MsnData msnData = new MsnData();
if (metaParser != null)
{
msnData.Meta.Parser = metaParser;
}
else
{
msnData.Meta.Parser = DefaultMetaParser;
}
if (labelParser != null)
{
msnData.Labels.Parser = labelParser;
}
else
{
msnData.Labels.Parser = DefaultLabelParser;
}
if (featureParser != null)
{
msnData.Feature.Parser = featureParser;
}
RankingTSVFile<MsnData> tsvFile = new RankingTSVFile<MsnData>(tsvFileName, msnData);
if (groupBoundary != null)
{
tsvFile.GroupBoundary = groupBoundary;
}
else
{
tsvFile.GroupBoundary = DefaultGroupBoundary;
}
return tsvFile;
}
public TsvFileLoader(RankingTSVFile<MsnData> tsvFile)
{
DataNullProc<MsnData> dataNullPro = new DataNullProc<MsnData>();
IDataEnum<MsnData, MsnData, DataNullProc<MsnData>> tsvDataEnum = new TsvDataStream<MsnData, MsnData, DataNullProc<MsnData>>(tsvFile, dataNullPro);
int gId = 0;
featureDataMatrix = new DataMatrixArray<float>();
List<float> labelList = new List<float>();
List<int> groupIdList = new List<int>();
foreach (MsnData d in tsvDataEnum)
{
if (featureNames == null)
{
featureNames = new string[d.Feature.NumColumns];
for (int i = 0; i < d.Feature.NumColumns; i++)
{
featureNames[i] = tsvFile.ColumnNames[d.Feature.Parser.columnIndex(i)];
}
}
//get labels
for (int i = 0; i < d.Labels.Data.NumRows; i++)
{
labelList.Add(d.Labels.Data.GetValue(i, 0));
groupIdList.Add(gId); // group index/Id
}
//get feature data
featureDataMatrix.Add(d.Feature.Data);
gId++;
}
int numRows = labelList.Count;
labels = new float[numRows];
groupId = new int[numRows];
for (int i = 0; i < numRows; i++)
{
labels[i] = labelList[i];
groupId[i] = groupIdList[i];
}
}