tools.cc

一种聚类算法,名字是cocluster

          if (*(argv+1) == NULL){
            cout << "  !!! Invalid col permutation number: (col permutation number should be specified.) !!!" << endl << endl;
            validCLA = false;
          }
          if (validCLA && !isdigit((*(argv+1))[0])){
            cout << "  !!! Invalid col permutation number: " << *(argv+1) << " !!!" << endl << endl;
            validCLA = false;
          }
	  tempNum = atoi(*(argv+1));
	  if (tempNum <= 0){
	    cout << "  !!! Invalid col permutation number: (col permutation number should be a positive integer.): " << tempNum << " !!!" << endl;
	    validCLA = false;
	  }
          if (validCLA){
	    myCLA.numColPermutation = atoi(*(++argv));
	    myCLA.colInitializationMethod = PERMUTATION_INIT;
	  }
	  break;
        case ROW_LABEL:
          if (*(argv+1) == NULL){
            cout << "  !!! Invalid row permutation number: (row permutation number should be specified.) !!!" << endl << endl;
            validCLA = false;
          }
          if (validCLA && !isdigit((*(argv+1))[0])){
            cout << "  !!! Invalid row permutation number: " << *(argv+1) << " !!!" << endl << endl;
            validCLA = false;
          }
          tempNum = atoi(*(argv+1));
	  if (tempNum <= 0){
	    cout << "  !!! Invalid row permutation number: (row permutation number should be a positive integer.): " << tempNum << " !!!" << endl;
	    validCLA = false;
	  }
          if (validCLA){
	    myCLA.numRowPermutation = atoi(*(++argv));
	    myCLA.rowInitializationMethod = PERMUTATION_INIT;
	  }
	  break;
        default:
	  cout << "  !!! Invalid permutation seeding label type: " << *argv << " !!!" << endl << endl;
          validCLA = false;
	  break;
      }
      break;
    default:
      cout << "  !!! Invalid seeding paramater(s): " << *argv << " !!!" << endl << endl;
      validCLA = false;
      break;
  }
  if (!validCLA){
    printSeedingType();
    myCLA.numInvalidCLA++;
  }
  return argv;
}


// get threshold value for batch update or local search from command-line... 
char **getThresholdInformation(int argc, char **argv, commandLineArgument &myCLA)
{
  bool validCLA = true;
  char tempThresholdInfo = toupper((*(++argv))[0]);
  switch (tempThresholdInfo){
    case BATCH_UPDATE_STEP:
    case LOCAL_SEARCH_STEP:
      break;
    default:
      cout << "  !!! Invalid threshold type: " << *argv << " !!!" << endl << endl;
      validCLA = false;
  }
  if (validCLA){ 
    if (*(argv+1) == NULL){
      cout << "  !!! Invalid row threshold: (row local search threshold should be specified.) !!!" << endl << endl;
      validCLA = false;
    }
    if (validCLA && (!isdigit((*(argv+1))[0]) && ((*(argv+1))[0] != '.'))){
      cout << "  !!! Invalid row threshold: " << *(argv+1) << " !!!" << endl << endl;
      validCLA = false;
    }
    if (validCLA && (*(argv+2) == NULL)){
      cout << "  !!! Invalid col threshold: (col threshold should be specified.) !!!" << endl << endl;
      validCLA = false;
    }
    if (validCLA && (!isdigit((*(argv+2))[0]) && ((*(argv+2))[0] != '.'))){
      cout << "  !!! Invalid col threshold: " << *(argv+2) << " !!!" << endl << endl;
      validCLA = false;
    }
    if (validCLA){
      switch (tempThresholdInfo){
        case BATCH_UPDATE_STEP:
          myCLA.rowBatchUpdateThreshold = atof(*(++argv));
	  myCLA.colBatchUpdateThreshold = atof(*(++argv));
        break;
        case LOCAL_SEARCH_STEP:
          myCLA.rowLocalSearchThreshold = atof(*(++argv));
          myCLA.colLocalSearchThreshold = atof(*(++argv));
          break;
        default:
          cout << "  !!! Invalid threshold type: " << *argv << " !!!" << endl << endl;
          validCLA = false;
          break;
      }
    }
  }
  if (!validCLA){
    printThresholdType();
    myCLA.numInvalidCLA++;
  }
  return argv;
}


// get type of batch update and local search from command-line...
char **getUpdateInformation(int argc, char **argv, commandLineArgument &myCLA)
{
  bool validCLA = true;
  switch (toupper((*(++argv))[0])){
    case BATCH_UPDATE_STEP:
      if (*(argv+1) == NULL){
        cout << "  !!! Invalid batch update type: (batch update type should be specified.) !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA && (!isdigit((*(argv+1))[0]))){
        cout << "  !!! Invalid batch update type: " << *(argv+1) << " !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA){
        switch ((*(++argv))[0]){
          case SINGLE_UPDATE:
	    myCLA.batchUpdateType = SINGLE_RESPECTIVELY;
	    break;
	  case SINGLE_SINGLE_UPDATE:
	    myCLA.batchUpdateType = SINGLE_IN_BATCH;
	    break;
          case MULTIPLE_UPDATE:
	    myCLA.batchUpdateType = MULTIPLE_RESPECTIVELY;
	    break;
          case SINGLE_FLIP:
	    myCLA.batchUpdateType = SINGLE_BY_FLIP;
	    break;
	  case MULTIPLE_FLIP:
	    myCLA.batchUpdateType = MULTIPLE_BY_FLIP;
	    break;
          default:
            cout << "  !!! Invalid batch update type: " << *argv << " !!!" << endl << endl;
            validCLA = false;
	    break;
        }
      }
      break;
    case LOCAL_SEARCH_STEP:
/*
      switch (toupper((*(++argv))[0])){
	case 'B':
          myCLA.localSearchType = BOTH_ROW_AND_COL;
	  break;
	case 'C':
          myCLA.localSearchType = COL_ONLY;
	  break;
	case 'R':
          myCLA.localSearchType = ROW_ONLY;
	  break;
        default:
          cout << "  Invalid local search update type: " << *argv << endl << endl;
          validCLA = false;
 	  break;
      }
*/
      if (*(argv+1) == NULL){
        cout << "  !!! Invalid row local search length: (row local search length should be specified.) !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA && (!isdigit((*(argv+1))[0]) && (*(argv+1))[0] != '-')){
        cout << "  !!! Invalid row local search length: " << *(argv+1) << " !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA && (*(argv+2) == NULL)){
        cout << "  !!! Invalid col local search length: (col local search length should be specified.) !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA && (!isdigit((*(argv+2))[0]) && (*(argv+2))[0] != '-')){
        cout << "  !!! Invalid col local search length: " << *(argv+2) << " !!!" << endl << endl;
        validCLA = false;
      }
      if (validCLA){
        myCLA.rowLocalSearchLength = atoi(*(++argv));
        myCLA.colLocalSearchLength = atoi(*(++argv));
      }
      break;
    default:  
      cout << "  !!! Invalid update type: " << *argv << " !!!" << endl << endl;
      validCLA = false;
      break;
  }
  if (!validCLA){
    printUpdateType();
    myCLA.numInvalidCLA++;
  }
  return argv;
}


// get information of taking reverse of rows from command-line...
char **getTakingReverse(int argc, char **argv, commandLineArgument &myCLA)
{
  bool validCLA = true;
  if (*(argv+1) == NULL){
    cout << "  !!! Invalid parameter: (0 or 1 should be specified.) !!!" << endl << endl;
    validCLA = false;
  }
  if (validCLA && !isdigit((*(argv+1))[0])){
    cout << "  !!! Invalid parameter: " << *(argv+1) << " !!!" << endl << endl;
    validCLA = false;
  }
  if (validCLA){
    switch (atoi(*(++argv))){
      case ON:
        myCLA.takingReverse = true;
	break;
      case OFF:
        myCLA.takingReverse = false;
	break;
      default:
        cout << " !!! Invalid parameter: (0 or 1 should be specified.): " << myCLA.numRun << " !!!" << endl << endl;
        validCLA = false;
	break;
    }
  }
  if (!validCLA){
    printTakingReverse();
    myCLA.numInvalidCLA++;
  }
  return argv;
}


// get co-clustering parameter values from command-line...
void getCommandLine(int argc, char **argv, commandLineArgument &myCLA)
{
  bool validCLA = true;
  for (argv++; *argv != NULL; argv++){        
    if ((*argv)[0] == '-'){
      switch (toupper((*argv)[1])){
        case MY_ALGORITHM:		// -A  algorithmType
          argv = getAlgorithmType(argc, argv, myCLA);
	  break;
        case MY_NUM_COL_CLUSTER:	// -C  numColCluster
          argv = getColClusterNum(argc, argv, myCLA);
          break;
        case MY_DUMP_LEVEL:		// -D  dumpLevel
          argv = getDumpLevel(argc, argv, myCLA);
          break;
        case MY_SHOWING_EACH_CLUSTER:	// -E  dumpLevel
          argv = getShowingEachCluster(argc, argv, myCLA);
          break;
        case MY_HELP:
	  printHelp();
	  break;
	case MY_INPUT_FILE:		// -I  inputMatrixType  inputFormatType  inputFilename
          argv = getInputFileInformation(argc, argv, myCLA);
          break;
        case MY_CLASS_FILE:		// -K  classLabelSelection  classOffsetType  classLabelFilename
	  argv = getClassFileInformation(argc, argv, myCLA);
	  break;
        case MY_SMOOTHING:		// -M smoothingType
          argv = getSmoothingType(argc, argv, myCLA);
          break;
        case MY_NUM_RUN:		// -N  runNum
          argv = getRunNum(argc, argv, myCLA);
          break;
        case MY_OUTPUT_FILE:		// -O  outputLabelType  outputAccessMode  labelOffsetType  outputFilename
          argv = getOutputFileInformation(argc, argv, myCLA);
          break;
        case MY_NUM_ROW_CLUSTER:	// -R  numRowCluster
          argv = getRowClusterNum(argc, argv, myCLA);
          break;
        case MY_SEEDING:		// -S  seedingType  seedingLabelSelection [labelOffsetType seedingFilename]
          argv = getSeedingInformation(argc, argv, myCLA);
          break;
        case MY_THRESHOLD:		// -T thresholdType thresholdValue
          argv = getThresholdInformation(argc, argv, myCLA);
          break;
        case MY_UPDATE:			// -U  updateType  updateOrderSelection
          argv = getUpdateInformation(argc, argv, myCLA);
          break;
        case MY_COMPUTING_ONE_WAY:	// -W
          argv = getComputingOneWayObjective(argc, argv, myCLA);
           break;
        case MY_TAKING_REVERSE:		// -X
          argv = getTakingReverse(argc, argv, myCLA);
          break;
        default:
//          printHelp();
	  cout << "  !!! Invalid argument setting: " << *argv << " !!!" << endl << endl;
          validCLA = false;
          break;
      }
    } else {
      sprintf(myCLA.inputFilename, "%s",*argv);
      if (strcmp(myCLA.coclusterFilename, "") == 0)
        extractFilename(*argv, myCLA.coclusterFilename);
      strcpy(myCLA.dumpFilename, myCLA.coclusterFilename);
      strcpy(myCLA.objectiveFilename, myCLA.coclusterFilename);
      myCLA.havingArgument = true;
    }
  }
  if (!validCLA){
//    printHelp();
    myCLA.numInvalidCLA++;
  }
}


void extractFilename(char *path, char *name)
{
  int length = strlen(path);
  for(int i = length-1; i >= 0; i--)
    if ((path[i] == '/') || (path[i] == '\\')){
      i++;
      for (int j = i; j < length; j++)
        name[j-i] = path[j];
      break;
    } else if (i == 0){
      for (int j = i; j < length; j++)
        name[j-i]=path[j];
      break;
    }
}


void makeFilename(char *filename, char *suffix, commandLineArgument &myCLA)
{
  char buffer[DEFAULT_STRING_LENGTH];
  char rowInit = ' ';
  char colInit = ' ';
  switch (myCLA.rowInitializationMethod){
    case RANDOM_INIT:  		rowInit = 'r';	break;
    case RANDOM_PERTURB_INIT:	rowInit = 'p';	break;
    case FARTHEST_INIT:		rowInit = 'f';	break;
    case PERTURBATION_INIT:	rowInit = 'p';	break;
    case SEEDING_INIT:		rowInit = 's';  break;
    default:
      cout << "  !!! Invalid row initialization: " << myCLA.rowInitializationMethod << " !!!" << endl << endl;
      exit(EXIT_FAILURE);
      break;
  }
  switch (myCLA.colInitializationMethod){
    case RANDOM_INIT:  		colInit = 'r';	break;
    case RANDOM_PERTURB_INIT:	colInit = 'p';	break;
    case FARTHEST_INIT:		colInit = 'f';	break;
    case PERTURBATION_INIT:	colInit = 'p';	break;
    case SEEDING_INIT:		colInit = 's';  break;
    default:  
      cout << "  !!! Invalid column initialization: " << myCLA.colInitializationMethod << " !!!" << endl << endl;
      exit(EXIT_FAILURE);
      break;
  }
  sprintf(buffer, "%s_%c%c_%d_%d", suffix, rowInit, colInit, myCLA.numRowCluster, myCLA.numColCluster);
  strcat(filename, buffer); 
}


int *readMatrix(char *filename, denseStruct *mat, int &numEmptyCol, int formatType, int matrixType)
  //read in a dense matrix; but since either rows are to be clustered or columns are to be clustered,
  //we use 'matrixType' to identify that.
  //for dense matrix we assume there is NO empty vector.
{
  std::ifstream dimFile;
  std::ifstream inputFile;
  char tempFilename[FILENAME_LENGTH];
  char whole_line[DEFAULT_STRING_LENGTH];
  int *emptyColId;
  
  switch (formatType){