tools.cc

一种聚类算法,名字是cocluster

    case DIM_MATRIX_SEPARATE_FORMAT:
      sprintf(tempFilename, "%s%s", filename, "_dim");
      dimFile.open(tempFilename, ios::in);
      if (!dimFile.is_open()){
        cout << "  !!! Dimension file open error: " << tempFilename << " !!!" << endl;
        exit(EXIT_FAILURE);
      }
      dimFile >> mat->numRow >> mat->numCol;
      dimFile.getline(whole_line, DEFAULT_STRING_LENGTH, '\n');
      inputFile.open(filename, ios::in);
      if (!inputFile.is_open()){
        cout << "  !!! Input file open error: " << filename << " !!!" << endl;
        exit(EXIT_FAILURE);
      }
      break;
    case DIM_MATRIX_TOGETHER_FORMAT:
      inputFile.open(filename, ios::in);
      if (!inputFile.is_open()){
        cout << "  !!! Input file open error: " << filename << " !!!" << endl;
        exit(EXIT_FAILURE);
      }
      inputFile >> mat->numRow >> mat->numCol;
      inputFile.getline(whole_line, DEFAULT_STRING_LENGTH, '\n');
      break;
    default:
      cout << "  !!! Incorrect input format type: " << formatType << " !!!" << endl;
      exit(EXIT_FAILURE);
      break;
  }
                
  //cout<<mat->numRow<<" "<< mat->numCol<<endl;
  mat->numVal = mat->numRow * mat->numCol;
  mat->value = new double *[mat->numRow];
  for (int i = 0; i < mat->numRow; i++)
    mat->value[i]= new double[mat->numCol];
  memoryUsed += (mat->numRow * mat->numCol) * sizeof(double);
  
  cout << "  Reading matrix file ... " << filename << endl;
  switch (matrixType){
    case DENSE_MATRIX:
      numEmptyCol = 0;
      for (int i = 0; i < mat->numRow; i++)
        for (int j = 0; j < mat->numCol; j++)
          inputFile >> mat->value[i][j];
      break;
    case DENSE_MATRIX_TRANS:				// not used...
      numEmptyCol = 0;
      for (int i = 0; i < mat->numCol; i++)
        for (int j=0; j < mat->numRow; j++)
          inputFile >> mat->value[j][i];
      break;
    default:
      break;
  }
  dimFile.close();
  inputFile.close();
  emptyColId = new int[numEmptyCol+1];
  emptyColId[0] = mat->numCol;
  return  emptyColId;
}


int *readMatrix(char *fname, sparseStruct *mat, int &numEmptyCol, int formatType, char *scalingType)
// read in a sparse matrix from files into 'mat' and return #empty vectors and an array of empty vector IDs
{
  char filename[FILENAME_LENGTH];
  //clock_t start_clock, finish_clock;
  sprintf(filename, "%s%s", fname, "_dim");
  std::ifstream dimFile(filename);
  if (dimFile == 0)
    cout << "  !!! File open error: " << filename << " !!!" << endl << endl;
  sprintf(filename, "%s%s", fname, "_row_ccs");
  std::ifstream rowPtrFile(filename);
  if (rowPtrFile == 0)
    cout << "  !!! File open error: " << filename << " !!!" << endl << endl;
  sprintf(filename, "%s%s", fname, "_col_ccs");
  std::ifstream colIdxFile(filename);
  if (colIdxFile == 0)
    cout << "  !!! File open error: " << filename << " !!!" << endl << endl;
  sprintf(filename, "%s%s", fname, "_");
  if (formatType == 0)
    sprintf(filename, "%s%s", filename, TXX_SCALING);
  else
    sprintf(filename, "%s%s", filename, TFN_SCALING);  
  sprintf(filename, "%s%s", filename, "_nz");
  std::ifstream valFile(filename);
  if (valFile == 0)
      cout << "  !!! File open error: " << filename << " !!!" << endl;
  if (dimFile == 0 || rowPtrFile == 0 || colIdxFile == 0 || valFile == 0){
    cout << "  !!! Matrix file " << fname << "_* has missing file(s) !!!" << endl;
    exit(EXIT_FAILURE);
  }
  //data.width(MAX_DESC_STR_LENGTH);
  //data >> mat->descString;
  //data >> mat->numCol >> mat->numRow >> mat->numVal;
  //start_clock = clock();
  cout << "  Reading dimension file ... " << endl;
  dimFile >> mat->numRow >> mat->numCol >> mat->numVal;
  dimFile.close();
  mat->colPtr = new int[mat->numCol+1];
  mat->rowIdx = new int[mat->numVal];
  mat->value = new double[mat->numVal];
  //space used for storing the sparse matrix is as follows
  memoryUsed += (mat->numCol + 1 + mat->numVal) * sizeof(int) + mat->numVal * sizeof(double);
  //it is necessary to handle empty vectors separately
  int pre = -1;
  int *tempEmptyColId = new int[mat->numCol], *emptyColId;
  numEmptyCol = 0;
  cout << "  Reading column pointer file ... " << endl;
  for (int i = 0; i < mat->numCol+1; i++){
    colIdxFile >> (mat->colPtr)[i];
    if ((mat->colPtr)[i] == pre){
      tempEmptyColId[numEmptyCol] = i - 1;
      numEmptyCol++;
    }
    pre = (mat->colPtr)[i];
  }
  colIdxFile.close();
  emptyColId = new int[numEmptyCol+1];
  for(int i = 0; i < numEmptyCol; i++)
    emptyColId[i] = tempEmptyColId[i];
  emptyColId[numEmptyCol] = mat->numCol;
  delete [] tempEmptyColId;
  cout << "  Reading row index file ... " << endl;
  for (int i = 0; i < mat->numVal; i++)
    rowPtrFile >> (mat->rowIdx)[i];
  rowPtrFile.close();
  cout << "  Reading non-zero value file ... " << endl;
  for (int i = 0; i < mat->numVal; i++)
    valFile >> (mat->value)[i];
  valFile.close();
  cout << endl;
  //finish_clock = clock();
  //cout << "Reading file time: " << (finish_clock - start_clock) / 1e6 << " seconds." << endl;
  if (numEmptyCol == 0)
    cout << "  !!! No empty col found !!!" << endl;
  else {
    cout << "  !!! " << numEmptyCol << " empty col(s) found !!!" << endl;
    for(int i = 0; i < numEmptyCol; i++)
      cout << emptyColId[i] << " ";
    cout << endl;
  }
  return  emptyColId;
}


/*
void myReadMatrix(commandLineArgument myCLA, sparseStruct &sparseCCS, sparseStruct &sparseCRS, denseStruct &denseMat)
{
  switch (myCLA.inputMatrixType){
    case DENSE_MATRIX:
    case DENSE_MATRIX_TRANS:	// not used...
      myCLA.emptyColId = readMatrix(myCLA.inputFilename, &denseMat, myCLA.numEmptyCol, myCLA.inputFormatType, myCLA.inputMatrixType);
      assert(((myCLA.numRowCluster <= denseMat.numRow) && (myCLA.numColCluster <= denseMat.numCol)) && (myCLA.numRowCluster != denseMat.numRow || myCLA.numColCluster != denseMat.numCol));
      assert((myCLA.rowLocalSearchLength < denseMat.numRow) && (myCLA.colLocalSearchLength < denseMat.numCol));
      memoryUsed += denseMat.numRow * denseMat.numCol * sizeof(double);
      break;
    case SPARSE_MATRIX:
    default:
      myCLA.emptyColId = readMatrix(myCLA.inputFilename, &sparseCCS, myCLA.numEmptyCol, myCLA.inputFormatType, myCLA.scalingType);
      assert(((myCLA.numRowCluster <= sparseCCS.numRow) && (myCLA.numColCluster <= sparseCCS.numCol)) && (myCLA.numRowCluster != sparseCCS.numRow || myCLA.numColCluster != sparseCCS.numCol));
      assert((myCLA.rowLocalSearchLength < sparseCCS.numRow) && (myCLA.colLocalSearchLength < sparseCCS.numCol));
      sparseCRS.colPtr = new int[sparseCCS.numRow+1];
      sparseCRS.rowIdx = new int[sparseCCS.numVal];
      sparseCRS.value = new double[sparseCCS.numVal];
      sparseCRS.numCol = sparseCCS.numRow;
      sparseCRS.numRow = sparseCCS.numCol;
      sparseCRS.numVal = sparseCCS.numVal;
      convertSparse2Sparse(sparseCCS.numCol, sparseCCS.numRow, sparseCCS.numVal, sparseCCS.colPtr, sparseCCS.rowIdx, sparseCCS.value, sparseCRS.colPtr, sparseCRS.rowIdx, sparseCRS.value); 
 //     checkConvertSparse2Sparse(sparseCCS.numCol, sparseCCS.numRow, sparseCCS.numVal, sparseCCS.colPtr, sparseCCS.rowIdx, sparseCCS.value, sparseCRS.colPtr, sparseCRS.rowIdx, sparseCRS.value); 
      memoryUsed += (sparseCCS.numCol + sparseCCS.numRow + 2 * sparseCCS.numVal) * sizeof(int) + 2 * sparseCCS.numVal * sizeof(double);
      break;
  }
}
*/


/*
void myPreprocessMatrix(commandLineArgument myCLA, sparseStruct &sparseCCS, sparseStruct &sparseCRS, denseStruct &denseMat, Matrix *myCCS, Matrix *myCRS)
{
  switch (myCLA.inputMatrixType){
    case DENSE_MATRIX:
      myCCS = new DenseMatrix(denseMat.numRow, denseMat.numCol, denseMat.value); 
      myCRS = myCCS;
      if (myCLA.algorithmType == INFORMATION_THEORETIC_CC)
        myCCS->preprocess();
      break;
    case SPARSE_MATRIX:
      myCCS = new SparseMatrix(sparseCCS.numRow, sparseCCS.numCol, sparseCCS.numVal, sparseCCS.value, sparseCCS.rowIdx, sparseCCS.colPtr);
      if (myCLA.algorithmType == INFORMATION_THEORETIC_CC)
        myCCS->preprocess();
      myCRS = new SparseMatrix(sparseCRS.numRow, sparseCRS.numCol, sparseCRS.numVal, sparseCRS.value, sparseCRS.rowIdx, sparseCRS.colPtr); 
      if (myCLA.algorithmType == INFORMATION_THEORETIC_CC)
        myCRS->preprocess();
      break;
    default:
      break;
  }  
}
*/


/*
void myConstructCoclustering(commandLineArgument myCLA, Matrix *myCCS, Matrix *myCRS, Coclustering *myCC)
{
  switch (myCLA.algorithmType){
    case INFORMATION_THEORETIC_CC:
      myCC = new Itcc(myCCS, myCRS, myCLA);
      myCC->classPrefix = ITCC_CLASS;
      break;
    case MINIMUM_SUM_SQUARE_RESIDUE_I_CC:
      myCC = new MssrIcc(myCCS, myCRS, myCLA);
      myCC->classPrefix = MSSRICC_CLASS;
      break;
    case MINIMUM_SUM_SQUARE_RESIDUE_II_CC:
      myCC = new MssrIIcc(myCCS, myCRS, myCLA);
      myCC->classPrefix = MSSRIICC_CLASS;
      break;
    default:
      myCC = new Itcc(myCCS, myCRS, myCLA);
      myCC->classPrefix = ITCC_CLASS;
      break;
  }
}
*/


// Read column class/cluster labels from the file whose first row contains num, 
// and each of the other line contains a label (i.e., one label in a row).
int readLabel(char *filename, int numData, int *label, int classOffsetType)
{
  char whole_line[DEFAULT_STRING_LENGTH];
  int maxLabel = 0, num, numClass = 0;
//  cout << "  Reading class label file ... " << filename << endl;
  std::ifstream labelFile(filename);

  if (!labelFile.is_open()){
    if (strcmp(filename, "") != 0){
      cout << "  !!! File open error: " << filename << " !!!" << endl;
      exit(EXIT_FAILURE);
    }
  }
  labelFile >> num;
  labelFile.getline(whole_line, DEFAULT_STRING_LENGTH, '\n');	// skip all the other columns in the first line
  if (num != numData){
    cout << "  !!! Incorrect number of class labels in: " << filename << " !!!" << endl;
    exit(EXIT_FAILURE);
  }
//  if (!isSilent)	// not used...
    cout << "  Reading class label file ... " << filename << endl;
  for (int i = 0; i < num; i++){
    labelFile >> label[i];
    labelFile.getline(whole_line, DEFAULT_STRING_LENGTH, '\n');
    if (label[i] > maxLabel)
      maxLabel = label[i];
  }
  labelFile.close();
  if (classOffsetType == START_FROM_0) 
    numClass = maxLabel + 1;
  else {
    numClass = maxLabel;
    for (int i = 0; i < num; i++)
      label[i]--;
  }
  return numClass;
}


// Read row or column class/cluster labels from the file, 
// where each line contains either row or column labels.
// So, if the file consists of many lines, # of lines == # of label sets.
void readLabel(char *filename, int numLabel, int *label, int &numClass, int classOffsetType)
{
//  cout << "  Reading class label file ... " << filename << endl;
  std::ifstream labelFile(filename);
  if (!labelFile.is_open()){
    if (strcmp(filename, "") != 0)
      cout << "  File open error: " << filename << endl;
    exit(EXIT_FAILURE);
  }
  readLabel(labelFile, numLabel, label, numClass, classOffsetType);

  labelFile.close();
}


// Read row or column class/cluster labels from the file stream, 
// where each line contains either row or column labels, 
// So, it the file consists of many lines, # of lines == # of label sets.
//(i.e., it may contains several sets of labels.)
void readLabel(istream &labelFile, int numLabel, int *label, int &numClass, int classOffsetType)
{
//  cout << "  Reading class label file ... " << filename << endl;
//  if (!isSilent)		// not used...
    cout << "  Reading class label file ... " << endl;  
  int maxLabel = 0;
  for (int i = 0; i < numLabel; i++){
    labelFile >> label[i];
    if (label[i] > maxLabel)
      maxLabel = label[i];
  }
  if (classOffsetType == START_FROM_0) 
    numClass = maxLabel + 1;
  else {
    numClass = maxLabel;
    for (int i = 0; i < numLabel; i++)
      label[i]--;
  }
}


// Read row and column class/cluster labels from the file consisting of two lines, 
// first line for row labels and second line for col labels.
void readLabel(char *filename, int numRow, int numCol, int *rowLabel, int *colLabel, int &numRowClass, int &numColClass, int classOffsetType)
{
//  cout << "  Reading class label file ... " << filename << endl;
  std::ifstream labelFile(filename);
  if (!labelFile.is_open()){
    if (strcmp(filename, "") != 0)
      cout << "  File open error: " << filename << endl;
    exit(EXIT_FAILURE);
  }
  readLabel(labelFile, numRow, numCol, rowLabel, colLabel, numRowClass, numColClass, classOffsetType);
// Otherwise, the following two lines are equivalent to the line above.
//  readLabel(labelFile, numRow, rowLabel, numRowClass, classOffsetType);
//  readLabel(labelFile, numCol, colLabel, numColClass, classOffsetType);
  labelFile.close();
}


// Read row and column class/cluster labels from the file stream, 
// where the first line for row labels and second line for col labels, 
// and so on. (i.e., it may contains several sets of labels.)
void readLabel(istream &labelFile, int numRow, int numCol, int *rowLabel, int *colLabel, int &numRowClass, int &numColClass, int classOffsetType)
{
//  cout << "  Reading class label file ... " << filename << endl;
//  if (!isSilent)		// not used...
    cout << "  Reading class label file ... " << endl;  
  int maxLabel = 0;
  for (int i = 0; i < numRow; i++){
    labelFile >> rowLabel[i];
    if (rowLabel[i] > maxLabel)
      maxLabel = rowLabel[i];
  }
  if (classOffsetType == START_FROM_0) 
    numRowClass = maxLabel + 1;
  else {
    numRowClass = maxLabel;
    for (int i = 0; i < numRow; i++)
      rowLabel[i]--;
  }
  maxLabel = 0;
  for (int i = 0; i < numCol; i++){
    labelFile >> colLabel[i];
    if (colLabel[i] > maxLabel)
      maxLabel = colLabel[i];
  }
  if (classOffsetType == START_FROM_0) 
    numColClass = maxLabel + 1;  
  else {
    numColClass = maxLabel;
    for (int i = 0; i < numCol; i++)
      colLabel[i]--;
  }
}



/*
 * Description: