int ArrayKCluster(int k, int nTrials, char method, char dist, int* NodeMap)
{ int ifound = 0;
double error;
int ok;
kcluster(k, _rows, _columns, _data, _mask,
_geneweight, 1, nTrials, method, dist, NodeMap, &error, &ifound);
ok = SetClusterIndex('a', k, NodeMap);
if (ok) return ifound;
return -1;
}
char* oph_ccluster_kcluster(UDF_INIT *initid, UDF_ARGS *args, char *result, unsigned long *length, char *is_null, char *error)
{
if(!initid->ptr){
*error=0;
*is_null=0;
initid->ptr=(char *)calloc(1,sizeof(oph_string));
if(!initid->ptr){
pmesg(1, __FILE__, __LINE__, "Error allocating result\n");
*length=0;
*is_null=1;
*error=1;
return NULL;
}
initid->extension = calloc(1,sizeof(oph_ccluster_kcluster_extra));
if(!initid->extension){
pmesg(1, __FILE__, __LINE__, "Error allocating extension\n");
*length=0;
*is_null=1;
*error=1;
return NULL;
}
oph_ccluster_kcluster_extra *extra = (oph_ccluster_kcluster_extra *) initid->extension;
extra->k = (int) *((long long*) args->args[3]); // set cluster number
extra->method = OPH_CCLUSTER_KCLUSTER_KMEANS; // default method
extra->level = OPH_CCLUSTER_KCLUSTER_ALL; // default level
extra->npass = 1; // default npass
extra->type = OPH_DOUBLE; // default input type
if (!strncasecmp(args->args[0],"OPH_INT",args->lengths[0])) extra->type = OPH_INT;
else if (!strncasecmp(args->args[0],"OPH_SHORT",args->lengths[0])) extra->type = OPH_SHORT;
else if (!strncasecmp(args->args[0],"OPH_BYTE",args->lengths[0])) extra->type = OPH_BYTE;
else if (!strncasecmp(args->args[0],"OPH_LONG",args->lengths[0])) extra->type = OPH_LONG;
else if (!strncasecmp(args->args[0],"OPH_FLOAT",args->lengths[0])) extra->type = OPH_FLOAT;
else if (!strncasecmp(args->args[0],"OPH_DOUBLE",args->lengths[0])) extra->type = OPH_DOUBLE;
else {
pmesg(1, __FILE__, __LINE__, "Invalid input data type!\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
int i;
for (i = 4; i < args->arg_count; i++) {
if (args->arg_type[i]==INT_RESULT) { // npass
extra->npass = (int) *((long long*) args->args[i]);
if (extra->npass < 1) {
pmesg(1, __FILE__, __LINE__, "npass must be >= 1!\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
} else if (args->arg_type[i]==STRING_RESULT) {
if (!strncasecmp(args->args[i],"KMEANS",args->lengths[i])) {
extra->method = OPH_CCLUSTER_KCLUSTER_KMEANS;
} else if (!strncasecmp(args->args[i],"KMEDIANS",args->lengths[i])) {
extra->method = OPH_CCLUSTER_KCLUSTER_KMEDIANS;
} else if (!strncasecmp(args->args[i],"CENTROIDS",args->lengths[i])) {
extra->level = OPH_CCLUSTER_KCLUSTER_CENTROIDS;
} else if (!strncasecmp(args->args[i],"LABELS",args->lengths[i])) {
extra->level = OPH_CCLUSTER_KCLUSTER_LABELS;
} else if (!strncasecmp(args->args[i],"ALL",args->lengths[i])) {
extra->level = OPH_CCLUSTER_KCLUSTER_ALL;
} else {
pmesg(1, __FILE__, __LINE__, "invalid argument %d!\n",i);
*length=0;
*is_null=0;
*error=1;
return NULL;
}
} else {
pmesg(1, __FILE__, __LINE__, "wrong type for argument %d!\n",i);
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
extra->npoints = args->lengths[2] / core_sizeof(extra->type);
extra->data = (double **) malloc(sizeof(double *) * extra->npoints);
if (!extra->data) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->npoints; i++) {
extra->data[i] = (double *) malloc(sizeof(double));
if (!extra->data[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
extra->mask = (int **) malloc(sizeof(int *) * extra->npoints);
if (!extra->mask) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->npoints; i++) {
extra->mask[i] = (int *) malloc(sizeof(int));
if (!extra->mask[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
extra->clusterid = (int *) malloc(sizeof(int) * extra->npoints);
if (!extra->clusterid) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
oph_stringPtr output = (oph_stringPtr) initid->ptr;
if (extra->level==OPH_CCLUSTER_KCLUSTER_CENTROIDS) {
extra->cdata = (double **) malloc(sizeof(double *) * extra->k);
if (!extra->cdata) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->k; i++) {
extra->cdata[i] = (double *) malloc(sizeof(double));
if (!extra->cdata[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
extra->cmask = (int **) malloc(sizeof(int *) * extra->k);
if (!extra->cmask) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->k; i++) {
extra->cmask[i] = (int *) malloc(sizeof(int));
if (!extra->cmask[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
unsigned long len = (unsigned long ) strlen("OPH_DOUBLE");
core_set_type(output, "OPH_DOUBLE" ,&len);
if(core_set_elemsize(output)){
pmesg(1, __FILE__, __LINE__, "Error on setting result elements size\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
output->length = (unsigned long *) malloc(sizeof(unsigned long));
if (!output->length) {
pmesg(1, __FILE__, __LINE__, "Error allocating length\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
*(output->length) = (unsigned long) extra->k * output->elemsize;
output->content = (char *) malloc(*(output->length));
if (!output->content) {
pmesg(1, __FILE__, __LINE__, "Error allocating output\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if(core_set_numelem(output)){
pmesg(1, __FILE__, __LINE__, "Error on counting result elements\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
} else if (extra->level==OPH_CCLUSTER_KCLUSTER_ALL) {
extra->cdata = (double **) malloc(sizeof(double *) * extra->k);
if (!extra->cdata) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->k; i++) {
extra->cdata[i] = (double *) malloc(sizeof(double));
if (!extra->cdata[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
extra->cmask = (int **) malloc(sizeof(int *) * extra->k);
if (!extra->cmask) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (i = 0; i < extra->k; i++) {
extra->cmask[i] = (int *) malloc(sizeof(int));
if (!extra->cmask[i]) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
unsigned long len = (unsigned long ) strlen("OPH_DOUBLE");
core_set_type(output, "OPH_DOUBLE" ,&len);
if(core_set_elemsize(output)){
pmesg(1, __FILE__, __LINE__, "Error on setting result elements size\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
output->length = (unsigned long *) malloc(sizeof(unsigned long));
if (!output->length) {
pmesg(1, __FILE__, __LINE__, "Error allocating length\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
*(output->length) = (unsigned long) extra->npoints * output->elemsize;
output->content = (char *) malloc(*(output->length));
if (!output->content) {
pmesg(1, __FILE__, __LINE__, "Error allocating output\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if(core_set_numelem(output)){
pmesg(1, __FILE__, __LINE__, "Error on counting result elements\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
} else {
unsigned long len = (unsigned long ) strlen("OPH_INT");
core_set_type(output, "OPH_INT" ,&len);
if(core_set_elemsize(output)){
pmesg(1, __FILE__, __LINE__, "Error on setting result elements size\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
output->length = (unsigned long *) malloc(sizeof(unsigned long));
if (!output->length) {
pmesg(1, __FILE__, __LINE__, "Error allocating length\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
*(output->length) = (unsigned long) extra->npoints * output->elemsize;
output->content = (char *) malloc(*(output->length));
if (!output->content) {
pmesg(1, __FILE__, __LINE__, "Error allocating output\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if(core_set_numelem(output)){
pmesg(1, __FILE__, __LINE__, "Error on counting result elements\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
}
}
if (*error)
{
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if (*is_null)
{
*length=0;
*is_null=1;
*error=0;
return NULL;
}
oph_stringPtr output = (oph_stringPtr) initid->ptr;
oph_ccluster_kcluster_extra *extra = (oph_ccluster_kcluster_extra *) initid->extension;
double weight[1] = {1.0};
char method = (extra->method==OPH_CCLUSTER_KCLUSTER_KMEANS)?'a':'m';
double err;
int ifound;
int n;
switch (extra->type) {
case OPH_INT:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = (double) ((int *)(args->args[2]))[n];
extra->mask[n][0] = 1;
}
break;
case OPH_SHORT:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = (double) ((short *)(args->args[2]))[n];
extra->mask[n][0] = 1;
}
break;
case OPH_BYTE:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = (double) ((char *)(args->args[2]))[n];
extra->mask[n][0] = 1;
}
break;
case OPH_LONG:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = (double) ((long long *)(args->args[2]))[n];
extra->mask[n][0] = 1;
}
break;
case OPH_FLOAT:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = (double) ((float *)(args->args[2]))[n];
if (isnan(extra->data[n][0])) {
extra->mask[n][0] = 0;
} else {
extra->mask[n][0] = 1;
}
}
break;
case OPH_DOUBLE:
for (n = 0; n < extra->npoints; n++) {
extra->data[n][0] = ((double *)(args->args[2]))[n];
if (isnan(extra->data[n][0])) {
extra->mask[n][0] = 0;
} else {
extra->mask[n][0] = 1;
}
}
break;
default:
pmesg(1, __FILE__, __LINE__, "Invalid input type\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
kcluster(extra->k,extra->npoints,1,extra->data,extra->mask,weight,0,extra->npass,method,'e',extra->clusterid,&err,&ifound);
if(ifound==0){
pmesg(1, __FILE__, __LINE__, "k > number of points!\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if(ifound==-1){
pmesg(1, __FILE__, __LINE__, "Error allocating memory for clustering\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
if (extra->level==OPH_CCLUSTER_KCLUSTER_CENTROIDS) {
if (!getclustercentroids(extra->k,extra->npoints,1,extra->data,extra->mask,extra->clusterid,extra->cdata,extra->cmask,0,method)) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory for centroids retrieval\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (n = 0; n < extra->k; n++) {
if (extra->cmask[n][0]==0) {
pmesg(1, __FILE__, __LINE__, "All cluster members are missing for cluster %d\n",n);
*length=0;
*is_null=0;
*error=1;
return NULL;
} else {
((double *)(output->content))[n] = extra->cdata[n][0];
}
}
} else if (extra->level==OPH_CCLUSTER_KCLUSTER_ALL) {
if (!getclustercentroids(extra->k,extra->npoints,1,extra->data,extra->mask,extra->clusterid,extra->cdata,extra->cmask,0,method)) {
pmesg(1, __FILE__, __LINE__, "Error allocating memory for centroids retrieval\n");
*length=0;
*is_null=0;
*error=1;
return NULL;
}
for (n = 0; n < extra->npoints; n++) {
if (extra->cmask[extra->clusterid[n]][0]==0) {
pmesg(1, __FILE__, __LINE__, "All cluster members are missing for cluster %d\n",extra->clusterid[n]);
*length=0;
*is_null=0;
*error=1;
return NULL;
} else {
((double *)(output->content))[n] = extra->cdata[extra->clusterid[n]][0];
}
}
} else {
memcpy(output->content,extra->clusterid,*(output->length));
}
*length=*(output->length);
return output->content;
}
/* @api private */
VALUE rb_do_kcluster(int argc, VALUE *argv, VALUE self) {
VALUE size, data, mask, weights, options;
rb_scan_args(argc, argv, "21", &size, &data, &options);
if (TYPE(data) != T_ARRAY)
rb_raise(rb_eArgError, "data should be an array of arrays");
if (NIL_P(size) || NUM2INT(rb_Integer(size)) > RARRAY_LEN(data))
rb_raise(rb_eArgError, "size should be > 0 and <= data size");
mask = get_value_option(options, "mask", Qnil);
if (!NIL_P(mask) && TYPE(mask) != T_ARRAY)
rb_raise(rb_eArgError, "mask should be an array of arrays");
int transpose = get_bool_option(options, "transpose", 0);
int npass = get_int_option(options, "iterations", DEFAULT_ITERATIONS);
// a = average, m = means
int method = get_int_option(options, "method", 'a');
// e = euclidian,
// b = city-block distance
// c = correlation
// a = absolute value of the correlation
// u = uncentered correlation
// x = absolute uncentered correlation
// s = spearman's rank correlation
// k = kendall's tau
int dist = get_int_option(options, "metric", 'e');
// initial assignment
int assign = get_int_option(options, "seed", 0);
int i,j;
int nrows = RARRAY_LEN(data);
int ncols = RARRAY_LEN(rb_ary_entry(data, 0));
int nsets = NUM2INT(rb_Integer(size));
double **cdata = (double**)malloc(sizeof(double*)*nrows);
int **cmask = (int **)malloc(sizeof(int *)*nrows);
double *cweights = (double *)malloc(sizeof(double )*ncols);
double **ccentroid;
int *ccluster, **ccentroid_mask, dimx = nrows, dimy = ncols, cdimx = nsets, cdimy = ncols;
for (i = 0; i < nrows; i++) {
cdata[i] = (double*)malloc(sizeof(double)*ncols);
cmask[i] = (int *)malloc(sizeof(int )*ncols);
for (j = 0; j < ncols; j++) {
cdata[i][j] = NUM2DBL(rb_Float(rb_ary_entry(rb_ary_entry(data, i), j)));
cmask[i][j] = NIL_P(mask) ? 1 : NUM2INT(rb_Integer(rb_ary_entry(rb_ary_entry(mask, i), j)));
}
}
weights = NIL_P(options) ? Qnil : rb_hash_aref(options, ID2SYM(rb_intern("weights")));
for (i = 0; i < ncols; i++) {
cweights[i] = NIL_P(weights) ? 1.0 : NUM2DBL(rb_Float(rb_ary_entry(weights, i)));
}
if (transpose) {
dimx = ncols;
dimy = nrows;
cdimx = nrows;
cdimy = nsets;
}
ccluster = (int *)malloc(sizeof(int )*dimx);
ccentroid = (double**)malloc(sizeof(double*)*cdimx);
ccentroid_mask = (int **)malloc(sizeof(int *)*cdimx);
for (i = 0; i < cdimx; i++) {
ccentroid[i] = (double*)malloc(sizeof(double)*cdimy);
ccentroid_mask[i] = (int *)malloc(sizeof(int )*cdimy);
}
int ifound;
double error;
kcluster(nsets,
nrows, ncols, cdata, cmask, cweights, transpose, npass, method, dist, ccluster, &error, &ifound, assign);
getclustercentroids(nsets,
nrows, ncols, cdata, cmask, ccluster, ccentroid, ccentroid_mask, transpose, method);
VALUE result = rb_hash_new();
VALUE cluster = rb_ary_new();
VALUE centroid = rb_ary_new();
for (i = 0; i < dimx; i++)
rb_ary_push(cluster, INT2NUM(ccluster[i]));
for (i = 0; i < cdimx; i++) {
VALUE point = rb_ary_new();
for (j = 0; j < cdimy; j++)
rb_ary_push(point, DBL2NUM(ccentroid[i][j]));
rb_ary_push(centroid, point);
}
rb_hash_aset(result, ID2SYM(rb_intern("cluster")), cluster);
rb_hash_aset(result, ID2SYM(rb_intern("centroid")), centroid);
rb_hash_aset(result, ID2SYM(rb_intern("error")), DBL2NUM(error));
rb_hash_aset(result, ID2SYM(rb_intern("repeated")), INT2NUM(ifound));
for (i = 0; i < nrows; i++) {
free(cdata[i]);
free(cmask[i]);
}
for (i = 0; i < cdimx; i++) {
free(ccentroid[i]);
free(ccentroid_mask[i]);
}
free(cdata);
free(cmask);
free(ccentroid);
free(ccentroid_mask);
free(cweights);
free(ccluster);
return result;
}
bool BagOfFeatures::buildKClustering(int numClusters, int pass, char method, char dist)
{
if(dictionary != NULL)
cvReleaseMat(&dictionary);
cout << "Initializing the data..." << endl;
int i, j;
int k = 0, l = 0, m = 0;
int size;
int totalImages = 0;
double* error = new double [numFeatures];
int ifound;
int *clusterID = new int [numFeatures];
double ** featureData = new double* [numFeatures];
// Allocate mask and set it all to 1 (assume no missing data)
int ** mask = new int* [numFeatures];
for(i = 0; i < numFeatures; i++)
{
mask[i] = new int [descrSize];
featureData[i] = new double [descrSize];
for(j = 0; j < descrSize; j++)
mask[i][j] = 1;
}
// Set the weights equal, all 1
double* weight = new double [descrSize];
for(i = 0; i < descrSize; i++)
weight[i] = 1.0;
// For each class
for(m = 0; m < numClasses; m++)
{
totalImages = data[m].getTrainSize();
// For each image in that class...
for(l = 0; l < totalImages; l++)
{
size = trainObject[m].featureSet[l].size;
// for each feature in that image...
for(i = 0; i < size; i++)
{
// Copy the descriptor into the data array
for(j = 0; j < descrSize; j++)
{
featureData[k][j] = (double)trainObject[m].featureSet[l].descriptors[i][j];
//cout << featureData[k][j] << " ";
}
//cout << endl;
k++;
}
}
}
cout << "Clustering data..." << endl;
kcluster(numClusters, numFeatures, descrSize, featureData,
mask, weight, 0, pass, method, dist,
clusterID, error, &ifound);
cout << "Computing cluster centers and building dictionary..." << endl;
int* indexCount = new int [numClusters];
int index;
float *ptrCenter;
dictionary = cvCreateMat(numClusters, descrSize, CV_32FC1);
cvSetZero(dictionary);
for(i = 0; i < numClusters; i++)
indexCount[i] = 0;
// Figure out how many clusters per index
for(i = 0; i < numFeatures; i++)
{
index = clusterID[i];
indexCount[index]++;
ptrCenter = (float *)(dictionary->data.ptr + index * dictionary->step);
for(j = 0; j < descrSize; j++)
{
ptrCenter[j] += (float)featureData[i][j];
}
}
for(i = 0; i < numClusters; i++)
{
ptrCenter = (float *)(dictionary->data.ptr + i * dictionary->step);
//cout << i << " \t\t\t" << indexCount[i] << endl << endl;
float t = indexCount[i];
for(j = 0; j < descrSize; j++)
{
ptrCenter[j] /= (float)indexCount[i];
}
}
// Release all memory
for(i = 0; i < numFeatures; i++)
{
delete [] mask[i];
delete [] featureData[i];
}
delete [] featureData;
delete [] mask;
delete [] weight;
delete [] indexCount;
delete [] error;
delete [] clusterID;
// Make sure that the tree was allocated
return true;
}
// calculate average comembership matrix in "resampNum" times of resampling
void calcAveComemMatrix(int nclusters, int nrow, int ncol, double** data, int** mask,
double weight[], int npass, int resampNum, double subSampPercnt, short** aveComemMatrix)
{
int subSampNum=int(nrow*subSampPercnt);
int i,j,k;
double** subData;
double** subCdata;
int** subCmask;
int** subMask;
int* subClusterid=(int*)malloc((size_t)subSampNum*sizeof(int));
int* allClusterid=(int*)malloc((size_t)nrow*sizeof(int));
long* originalIndex=(long*)malloc((size_t)nrow*sizeof(int));
double error; int ifound;
subData = (double**)malloc((size_t)subSampNum*sizeof(double*));
for(i=0; i<subSampNum; i++)
subData[i] = (double*)malloc((size_t)ncol*sizeof(double));
subCdata = (double**)malloc((size_t)nclusters*sizeof(double*));
for(i=0; i<nclusters; i++)
subCdata[i] = (double*)malloc((size_t)ncol*sizeof(double));
subMask= (int**)malloc((size_t)subSampNum*sizeof(int*));
for(i=0; i<subSampNum; i++)
subMask[i] = (int*)malloc((size_t)ncol*sizeof(int));
subCmask= (int**)malloc((size_t)nclusters*sizeof(int*));
for(i=0; i<nclusters; i++)
subCmask[i] = (int*)malloc((size_t)ncol*sizeof(int));
for(i=0; i<nrow; i++)
for(j=0; j<nrow; j++)
aveComemMatrix[i][j]=0;
for(i=0; i<nrow; i++)
originalIndex[i]=i;
for(k=0; k<resampNum; k++)
{
//printf("the %dth resampling!\n", k);
genprm(originalIndex, nrow);
for(i=0; i<subSampNum; i++)
{
for(j=0; j<ncol; j++)
{
subData[i][j]=data[originalIndex[i]][j];
subMask[i][j]=mask[originalIndex[i]][j];
}
}
if(nclusters<subSampNum)
kcluster(nclusters, subSampNum, ncol, subData, subMask, weight, 0, npass, 'a', 'e', subClusterid, subCdata, subCmask, &error, &ifound);
else {printf("Error: nclusters larger than subSampNum!!"); scanf("%d", &i);}
for(i=0; i<nrow; i++)
allClusterid[i]=which_minDist(ncol, data, subCdata, mask, subCmask, weight, i, nclusters);
for(i=0; i<nrow; i++)
{
for(j=0; j<nrow; j++)
{
if(allClusterid[i]==allClusterid[j])
aveComemMatrix[i][j] += 1;
}
}
}
//free memory
for(i=0; i<subSampNum; i++) free(subData[i]);
for(i=0; i<subSampNum; i++) free(subMask[i]);
for(i=0; i<nclusters; i++) free(subCdata[i]);
for(i=0; i<nclusters; i++) free(subCmask[i]);
free(subData);
free(subCdata);
free(subMask);
free(subCmask);
free(subClusterid);
free(allClusterid);
free(originalIndex);
}
void example_kmeans( int nrows, int ncols,
double** data,
int nclusters, int npass,
char dist, char* jobname,
int *clusterid)
/* Perform k-means clustering on genes */
{
int i, j, ii, nl, nc;
//const int nclusters = 3;
const int transpose = 0;
//const char dist = 'e';
const char method = 'a';
/* For method=='a', the centroid is defined as the mean over all elements
belonging to a cluster for each dimension.
For method=='m', the centroid is defined as the median over all elements
belonging to a cluster for each dimension.
*/
//int npass = 1;
int ifound = 0;
int test=0;
double error;
double distance;
int** index;
int* count;
double* weight = malloc(ncols*sizeof(double));
double** cdata = malloc(nclusters*sizeof(double*));
int** cmask = malloc(nclusters*sizeof(int*));
int** mask = NULL;
int n=0;
char* filename;
char* filename2;
char* filename3;
FILE *out1=NULL;
FILE *out2=NULL;
FILE *out3=NULL;
for (i = 0; i < nclusters; i++)
{ cdata[i] = malloc(ncols*sizeof(double));
cmask[i] = malloc(ncols*sizeof(int));
}
for (i = 0; i < ncols; i++) weight[i] = 1.0;
mask = (int **)calloc(sizeof(int*), nrows);
for (ii=0;ii<nrows;++ii) {
mask[ii] = (int *)calloc(sizeof(int),ncols);
}
for (nl=0; nl<nrows; ++nl) {
for (nc=0; nc<ncols; ++nc) {
mask[nl][nc] = 1;
}
}
n = 1 + strlen(jobname) + strlen("_K_G") + strlen(".ext");
if (dist)
{ int dummy = nclusters;
do n++; while (dummy/=10);
}
//avovk
printf("a je u omari :) \n");
filename = (char *)malloc(n*sizeof(char));
filename2 = (char *)malloc(n*sizeof(char));
filename3 = (char *)malloc(n*sizeof(char));
sprintf (filename, "%s_K_G%d.kgg", jobname, nclusters);
out1 = fopen( filename, "w" );
sprintf (filename2, "%s_K_G%d.dis", jobname, nclusters);
out2 = fopen( filename2, "w" );
sprintf (filename3, "%s_K_G%d.cen", jobname, nclusters);
out3 = fopen( filename3, "w" );
printf("======================== k-means clustering"
" ========================\n");
printf ("\n");
printf ("----- doing %d passes... go stretch your legs...\n",npass);
//npass = 3;
/* ANDREJ: This function returns different answers each time it is
executed. Does the library provide for ways to initialize the
random number generators used for the searching and initializations? */
kcluster( nclusters,
nrows,ncols,data,
mask,weight,
transpose,npass,
method,dist,
clusterid, &error, &ifound);
printf ("Solution found %d times; ", ifound);
printf ("within-cluster sum of distances is %f\n", error);
printf ("------- writing Cluster assignments to file:\t\t"
" %s_K_G%d.kgg\n",jobname, nclusters);
for (i = 0; i < nrows; i++)
fprintf (out1, "%09d\t %d\n", i, clusterid[i]);
fclose(out1); out1=NULL;
printf ("------- writing Distance between clusters to file:\t "
"%s_K_G%d.dis \n", jobname, nclusters);
fprintf (out2,"------- Distance between clusters:\n");
index = (int **)malloc(nclusters*sizeof(int*));
count = (int *)malloc(nclusters*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++) count[clusterid[i]]++;
for (i = 0; i < nclusters; i++) index[i] = malloc(count[i]*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++){
int id = clusterid[i];
index[id][count[id]] = i;
count[id]++;
}
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[0], count[1],
index[0], index[1], 'e', 'a', 0);
fprintf(out2,"Distance between 0 and 1: %7.3f\n", distance);
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[0], count[2],
index[0], index[2], 'e', 'a', 0);
fprintf(out2,"Distance between 0 and 2: %7.3f\n", distance);
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[1], count[2],
index[1], index[2], 'e', 'a', 0);
fprintf(out2,"Distance between 1 and 2: %7.3f\n", distance);
fclose(out2); out2=NULL;
printf ("------- writing Cluster centroids to file:\t\t"
"%s_K_G%d.cen\n",jobname, nclusters);
fprintf (out3,"------- Cluster centroids:\n");
getclustercentroids(nclusters, nrows, ncols, data, mask, clusterid,
cdata, cmask, 0, 'a');
fprintf(out3," coefficients:");
for(i=0; i<ncols; i++) fprintf(out3,"\t%7d", i);
fprintf(out3,"\n");
for (i = 0; i < nclusters; i++){
fprintf(out3,"Cluster %2d:", i);
for (j = 0; j < ncols; j++) fprintf(out3,"\t%7.3f", cdata[i][j]);
fprintf(out3,"\n");
}
fclose(out3); out3=NULL;
printf("Done...\n");
for (i = 0; i < nclusters; i++) free(index[i]);
free(index);
free(count);
for (i = 0; i < nclusters; i++){
free(cdata[i]);
free(cmask[i]);
}
for (ii=0;ii<nrows;++ii) {
if (mask[ii]) free(mask[ii]);
}
free(cdata);
free(cmask);
free(weight);
free(mask);
return;
}
void example_kmeans( int nrows, int ncols,
double** data,
int nclusters, int npass,
char dist, char* jobname)
/* Perform k-means clustering on genes */
{
int i, j, ii, nl, nc;
//const int nclusters = 3;
const int transpose = 0;
//const char dist = 'e';
const char method = 'a';
/* For method=='a', the centroid is defined as the mean over all elements
belonging to a cluster for each dimension.
For method=='m', the centroid is defined as the median over all elements
belonging to a cluster for each dimension.
*/
//int npass = 1;
int ifound = 0;
int test=0;
double error;
double distance;
int** index;
int* count;
double* weight = malloc(ncols*sizeof(double));
int* clusterid = malloc(nrows*sizeof(int));
double** cdata = malloc(nclusters*sizeof(double*));
int n=0;
char* filename;
char* filename2;
char* filename3;
FILE *out1=NULL;
FILE *out2=NULL;
FILE *out3=NULL;
for (i = 0; i < nclusters; i++)
{ cdata[i] = malloc(ncols*sizeof(double));
}
for (i = 0; i < ncols; i++) weight[i] = 1.0;
n = 1 + strlen(jobname) + strlen("_K_G") + strlen(".ext");
if (dist)
{ int dummy = nclusters;
do n++; while (dummy/=10);
}
//avovk
printf("a je u omari :) \n");
filename = malloc(n*sizeof(char));
filename2 = malloc(n*sizeof(char));
filename3 = malloc(n*sizeof(char));
sprintf (filename, "%s_K_G%d.kgg", jobname, nclusters);
out1 = fopen( filename, "w" );
sprintf (filename2, "%s_K_G%d.dis", jobname, nclusters);
out2 = fopen( filename2, "w" );
sprintf (filename3, "%s_K_G%d.cen", jobname, nclusters);
out3 = fopen( filename3, "w" );
printf("======================== k-means clustering"
" ========================\n");
printf ("\n");
printf ("----- doing %d passes... go stretch your legs...\n",npass);
//npass = 3;
kcluster(nclusters,nrows,ncols,data,weight,transpose,npass,method,dist,
clusterid, &error, &ifound);
printf ("Solution found %d times; ", ifound);
printf ("within-cluster sum of distances is %f\n", error);
printf ("------- writing Cluster assignments to file:\t\t"
" %s_K_G%d.kgg\n",jobname, nclusters);
for (i = 0; i < nrows; i++)
fprintf (out1, "%09d\t %d\n", i, clusterid[i]);
fclose(out1); out1=NULL;
printf ("------- writing Distance between clusters to file:\t %s_K_G%d.dis \n", jobname, nclusters);
fprintf (out2,"------- Distance between clusters:\n");
index = malloc(nclusters*sizeof(int*));
count = malloc(nclusters*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++) count[clusterid[i]]++;
for (i = 0; i < nclusters; i++) index[i] = malloc(count[i]*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++)
{ int id = clusterid[i];
index[id][count[id]] = i;
count[id]++;
}
for (i = 0; i < nclusters-1; i++)
{
for (j = 1+i; j < nclusters; j++)
{
distance = clusterdistance(nrows, ncols, data, weight, count[i], count[j], index[i], index[j], 'e', 'a', 0);
fprintf(out2,"Distance between %d and %d: %7.3f\n", i, j, distance);
// fprintf(stderr,"Distance between %d and %d: %7.3f\n", i, j, distance);
}
}
fclose(out2); out2=NULL;
printf ("------- writing Cluster centroids to file:\t\t%s_K_G%d.cen\n",jobname, nclusters);
fprintf (out3,"------- Cluster centroids:\n");
getclustercentroids(nclusters, nrows, ncols, data, clusterid, cdata, 0, 'a');
fprintf(out3," coefficients:");
for(i=0; i<ncols; i++) fprintf(out3,"\t%7d", i);
fprintf(out3,"\n");
for (i = 0; i < nclusters; i++){
fprintf(out3,"Cluster %2d:", i);
for (j = 0; j < ncols; j++) fprintf(out3,"\t%7.3f", cdata[i][j]);
fprintf(out3,"\n");
}
fclose(out3); out3=NULL;
printf("Done...\n");
/* call function to calculate distance between each voxel and centroid */
/* we will need:
count - number of elements in cluster as we allready have it
cdata - cluster centroids
clusterid
data */
getvoxlclusterdist(count, cdata, clusterid, data, jobname,
nclusters, nrows, ncols);
for (i = 0; i < nclusters; i++) free(index[i]);
free(index);
free(count);
for (i = 0; i < nclusters; i++){
free(cdata[i]);
}
free(cdata);
free(clusterid);
free(weight);
return;
}
void example_kmeans(int nrows, int ncols, double** data, int** mask)
/* Perform k-means clustering on genes */
{ int i, j;
const int nclusters = 3;
const int transpose = 0;
const char dist = 'e';
const char method = 'a';
int npass = 1;
int ifound = 0;
double error;
double distance;
int** index;
int* count;
double* weight = malloc(ncols*sizeof(double));
int* clusterid = malloc(nrows*sizeof(int));
double** cdata = malloc(nclusters*sizeof(double*));
int** cmask = malloc(nclusters*sizeof(int*));
for (i = 0; i < nclusters; i++)
{ cdata[i] = malloc(ncols*sizeof(double));
cmask[i] = malloc(ncols*sizeof(int));
}
for (i = 0; i < ncols; i++) weight[i] = 1.0;
printf("======================== k-means clustering ====================\n");
printf("\n");
printf("----- one pass of the EM algorithm (results may change)\n");
kcluster(nclusters,nrows,ncols,data,mask,weight,transpose,npass,method,dist,
clusterid, &error, &ifound);
printf ("Solution found %d times; within-cluster sum of distances is %f\n",
ifound, error);
printf ("Cluster assignments:\n");
for (i = 0; i < nrows; i++)
printf ("Gene %d: cluster %d\n", i, clusterid[i]);
printf ("\n");
printf("----- 1000 passes of the EM algorithm (result should not change)\n");
npass = 1000;
kcluster(nclusters,nrows,ncols,data,mask,weight,transpose,npass,method,dist,
clusterid, &error, &ifound);
printf ("Solution found %d times; ", ifound);
printf ("within-cluster sum of distances is %f\n", error);
printf ("Cluster assignments:\n");
for (i = 0; i < nrows; i++)
printf ("Gene %d: cluster %d\n", i, clusterid[i]);
printf ("\n");
printf ("------- Distance between clusters:\n");
index = malloc(nclusters*sizeof(int*));
count = malloc(nclusters*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++) count[clusterid[i]]++;
for (i = 0; i < nclusters; i++) index[i] = malloc(count[i]*sizeof(int));
for (i = 0; i < nclusters; i++) count[i] = 0;
for (i = 0; i < nrows; i++)
{ int id = clusterid[i];
index[id][count[id]] = i;
count[id]++;
}
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[0], count[1],
index[0], index[1], 'e', 'a', 0);
printf("Distance between 0 and 1: %7.3f\n", distance);
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[0], count[2],
index[0], index[2], 'e', 'a', 0);
printf("Distance between 0 and 2: %7.3f\n", distance);
distance =
clusterdistance(nrows, ncols, data, mask, weight, count[1], count[2],
index[1], index[2], 'e', 'a', 0);
printf("Distance between 1 and 2: %7.3f\n", distance);
printf ("\n");
printf ("------- Cluster centroids:\n");
getclustermean(nclusters, nrows, ncols, data, mask, clusterid,
cdata, cmask, 0);
printf(" Microarray:");
for(i=0; i<ncols; i++) printf("\t%7d", i);
printf("\n");
for (i = 0; i < nclusters; i++)
{ printf("Cluster %2d:", i);
for (j = 0; j < ncols; j++) printf("\t%7.3f", cdata[i][j]);
printf("\n");
}
printf("\n");
for (i = 0; i < nclusters; i++) free(index[i]);
free(index);
free(count);
for (i = 0; i < nclusters; i++)
{ free(cdata[i]);
free(cmask[i]);
}
free(cdata);
free(cmask);
free(clusterid);
free(weight);
return;
}
