Chiebot-Mirror
/
opencv
mirror of https://github.com/opencv/opencv.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

made random generators of MLL classes depended on default rng (theRNG) (#205).

Browse Source
pull/13383/head
Maria Dimashova 14 years ago
parent 7860d52e14
commit c104cdce96
10 changed files with 53 additions and 55 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 10
      modules/ml/include/opencv2/ml/ml.hpp
  2. 12
      modules/ml/src/ann_mlp.cpp
  3. 6
      modules/ml/src/data.cpp
  4. 8
      modules/ml/src/em.cpp
  5. 25
      modules/ml/src/ertrees.cpp
  6. 6
      modules/ml/src/gbt.cpp
  7. 12
      modules/ml/src/rtrees.cpp
  8. 6
      modules/ml/src/svm.cpp
  9. 21
      modules/ml/src/tree.cpp
  10. 2
      modules/traincascade/boost.cpp

10
modules/ml/include/opencv2/ml/ml.hpp
Unescape View File

@ -913,7 +913,7 @@ struct CV_EXPORTS CvDTreeTrainData
CvSet* cv_heap;
CvSet* nv_heap;
CvRNG rng;
cv::RNG* rng;
};
class CvDTree;
@ -1147,7 +1147,7 @@ protected:
CvMat* var_importance;
int nsamples;
CvRNG rng;
cv::RNG* rng;
CvMat* active_var_mask;
};
@ -1908,7 +1908,7 @@ protected:
CvMat* missing;
CvMat* class_labels;
CvRNG rng;
cv::RNG* rng;
int class_count;
float delta;
@ -2034,7 +2034,7 @@ protected:
int activ_func;
int max_count, max_buf_sz;
CvANN_MLP_TrainParams params;
CvRNG rng;
cv::RNG* rng;
};
/****************************************************************************************\
@ -2177,7 +2177,7 @@ protected:
CvMat* test_sample_idx;
int* sample_idx; // data of train_sample_idx and test_sample_idx
CvRNG rng;
cv::RNG* rng;
};

12
modules/ml/src/ann_mlp.cpp
Unescape View File

@ -95,7 +95,7 @@ CvANN_MLP::CvANN_MLP()
layer_sizes = wbuf = 0;
min_val = max_val = min_val1 = max_val1 = 0.;
weights = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
default_model_name = "my_nn";
clear();
}
@ -108,7 +108,7 @@ CvANN_MLP::CvANN_MLP( const CvMat* _layer_sizes,
layer_sizes = wbuf = 0;
min_val = max_val = min_val1 = max_val1 = 0.;
weights = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
default_model_name = "my_nn";
create( _layer_sizes, _activ_func, _f_param1, _f_param2 );
}
@ -190,7 +190,7 @@ void CvANN_MLP::init_weights()
double s = 0;
for( k = 0; k <= n1; k++ )
{
val = cvRandReal(&rng)*2-1.;
val = rng->uniform(0., 1.)*2-1.;
w[k*n2 + j] = val;
s += fabs(val);
}
@ -928,8 +928,8 @@ int CvANN_MLP::train_backprop( CvVectors x0, CvVectors u, const double* sw )
for( i = 0; i < count; i++ )
{
int tt;
j = (unsigned)cvRandInt(&rng) % count;
k = (unsigned)cvRandInt(&rng) % count;
j = (*rng)(count);
k = (*rng)(count);
CV_SWAP( _idx->data.i[j], _idx->data.i[k], tt );
}
}
@ -1507,7 +1507,7 @@ CvANN_MLP::CvANN_MLP( const Mat& _layer_sizes, int _activ_func,
layer_sizes = wbuf = 0;
min_val = max_val = min_val1 = max_val1 = 0.;
weights = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
default_model_name = "my_nn";
create( _layer_sizes, _activ_func, _f_param1, _f_param2 );
}

6
modules/ml/src/data.cpp
Unescape View File

@ -84,7 +84,7 @@ CvMLData :: CvMLData()
//flt_separator = '.';
class_map = new std::map<std::string, int>();
rng = cvRNG( -cvGetTickCount() );
rng = &cv::theRNG();
}
CvMLData :: ~CvMLData()
@ -608,8 +608,8 @@ void CvMLData :: mix_train_and_test_idx()
int n = values->rows;
for (int i = 0; i < n; i++)
{
int a = cvRandInt( &rng ) % n;
int b = cvRandInt( &rng ) % n;
int a = (*rng)(n);
int b = (*rng)(n);
int t;
CV_SWAP( sample_idx[a], sample_idx[b], t );
}

8
modules/ml/src/em.cpp
Unescape View File

@ -581,7 +581,7 @@ void CvEM::kmeans( const CvVectors& train_data, int nclusters, CvMat* labels,
__BEGIN__;
CvRNG rng = cvRNG(-1);
cv::RNG* rng = &cv::theRNG();
int i, j, k, nsamples, dims;
int iter = 0;
double max_dist = DBL_MAX;
@ -605,7 +605,7 @@ void CvEM::kmeans( const CvVectors& train_data, int nclusters, CvMat* labels,
{
for( i = 0; i < nsamples; i++ )
labels->data.i[i] = i*nclusters/nsamples;
cvRandShuffle( labels, &rng );
cvRandShuffle( labels, &rng->state );
}
for( ;; )
@ -702,7 +702,7 @@ void CvEM::kmeans( const CvVectors& train_data, int nclusters, CvMat* labels,
const float* s;
for( j = 0; j < 10; j++ )
{
i = cvRandInt( &rng ) % nsamples;
i = (*rng)(nsamples);
if( counters->data.i[labels->data.i[i]] > 1 )
break;
}
@ -738,7 +738,7 @@ void CvEM::kmeans( const CvVectors& train_data, int nclusters, CvMat* labels,
if( counters->data.i[k] == 0 )
for(;;)
{
i = cvRandInt(&rng) % nsamples;
i = (*rng)(nsamples);
j = labels->data.i[i];
if( counters->data.i[j] > 1 )
{

25
modules/ml/src/ertrees.cpp
Unescape View File

@ -91,7 +91,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
clear();
var_all = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
CV_CALL( set_params( _params ));
@ -444,7 +444,6 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
{
unsigned short* udst = 0;
int* idst = 0;
CvRNG* r = &rng;
if (is_buf_16u)
{
@ -457,8 +456,8 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
for( i = 0; i < sample_count; i++ )
{
int a = cvRandInt(r) % sample_count;
int b = cvRandInt(r) % sample_count;
int a = (*rng)(sample_count);
int b = (*rng)(sample_count);
unsigned short unsh = (unsigned short)vi;
CV_SWAP( udst[a], udst[b], unsh );
}
@ -474,8 +473,8 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
for( i = 0; i < sample_count; i++ )
{
int a = cvRandInt(r) % sample_count;
int b = cvRandInt(r) % sample_count;
int a = (*rng)(sample_count);
int b = (*rng)(sample_count);
CV_SWAP( idst[a], idst[b], vi );
}
}
@ -894,8 +893,8 @@ CvDTreeSplit* CvForestERTree::find_split_ord_class( CvDTreeNode* node, int vi, f
if (fdiff > epsilon)
{
is_find_split = true;
CvRNG* rng = &data->rng;
split_val = pmin + cvRandReal(rng) * fdiff ;
cv::RNG* rng = data->rng;
split_val = pmin + rng->uniform(0.f, 1.f) * fdiff ;
if (split_val - pmin <= FLT_EPSILON)
split_val = pmin + split_delta;
if (pmax - split_val <= FLT_EPSILON)
@ -1047,7 +1046,7 @@ CvDTreeSplit* CvForestERTree::find_split_cat_class( CvDTreeNode* node, int vi, f
for (int i = 0; i < valid_ccount; i++)
{
uchar temp;
int i1 = cvRandInt( rng ) % valid_ccount;
int i1 = cvRandInt( rng ) % valid_ccount;
int i2 = cvRandInt( rng ) % valid_ccount;
CV_SWAP( var_class_mask->data.ptr[i1], var_class_mask->data.ptr[i2], temp );
}
@ -1189,8 +1188,8 @@ CvDTreeSplit* CvForestERTree::find_split_ord_reg( CvDTreeNode* node, int vi, flo
if (fdiff > epsilon)
{
is_find_split = true;
CvRNG* rng = &data->rng;
split_val = pmin + cvRandReal(rng) * fdiff ;
cv::RNG* rng = data->rng;
split_val = pmin + rng->uniform(0.f, 1.f) * fdiff ;
if (split_val - pmin <= FLT_EPSILON)
split_val = pmin + split_delta;
if (pmax - split_val <= FLT_EPSILON)
@ -1745,8 +1744,8 @@ bool CvERTrees::grow_forest( const CvTermCriteria term_crit )
int i1, i2;
float temp;
i1 = cvRandInt( &rng ) % nsamples;
i2 = cvRandInt( &rng ) % nsamples;
i1 = (*rng)(nsamples);
i2 = (*rng)(nsamples);
CV_SWAP( mth_var_ptr[i1*dims], mth_var_ptr[i2*dims], temp );
// turn values of (m-1)-th variable, that were permuted

6
modules/ml/src/gbt.cpp
Unescape View File

@ -268,7 +268,7 @@ CvGBTrees::train( const CvMat* _train_data, int _tflag,
}
// subsample params and data
rng = CvRNG(time(0));
rng = &cv::theRNG();
int samples_count = get_len(sample_idx);
@ -698,8 +698,8 @@ void CvGBTrees::do_subsample()
if (subsample_test)
for (int i = 0; i < n; i++)
{
int a = cvRandInt( &rng ) % n;
int b = cvRandInt( &rng ) % n;
int a = (*rng)(n);
int b = (*rng)(n);
int t;
CV_SWAP( idx[a], idx[b], t );
}

12
modules/ml/src/rtrees.cpp
Unescape View File

@ -200,7 +200,7 @@ CvRTrees::CvRTrees()
data = NULL;
active_var_mask = NULL;
var_importance = NULL;
rng = cvRNG(0xffffffff);
rng = &cv::theRNG();
default_model_name = "my_random_trees";
}
@ -235,7 +235,7 @@ CvMat* CvRTrees::get_active_var_mask()
CvRNG* CvRTrees::get_rng()
{
return &rng;
return &rng->state;
}
bool CvRTrees::train( const CvMat* _train_data, int _tflag,
@ -375,7 +375,7 @@ bool CvRTrees::grow_forest( const CvTermCriteria term_crit )
cvZero( sample_idx_mask_for_tree );
for(i = 0; i < nsamples; i++ ) //form sample for creation one tree
{
int idx = cvRandInt( &rng ) % nsamples;
int idx = (*rng)(nsamples);
sample_idx_for_tree->data.i[i] = idx;
sample_idx_mask_for_tree->data.ptr[idx] = 0xFF;
}
@ -458,8 +458,8 @@ bool CvRTrees::grow_forest( const CvTermCriteria term_crit )
if( sample_idx_mask_for_tree->data.ptr[i] ) //the sample is not OOB
continue;
i1 = cvRandInt( &rng ) % nsamples;
i2 = cvRandInt( &rng ) % nsamples;
i1 = (*rng)(nsamples);
i2 = (*rng)(nsamples);
CV_SWAP( mth_var_ptr[i1*dims], mth_var_ptr[i2*dims], temp );
// turn values of (m-1)-th variable, that were permuted
@ -762,7 +762,7 @@ void CvRTrees::read( CvFileStorage* fs, CvFileNode* fnode )
CV_Error( CV_StsParseError, "Some <nclasses>, <nsamples>, <var_count>, "
"<nactive_vars>, <oob_error>, <ntrees> of tags are missing" );
rng = CvRNG( -1 );
rng = &cv::theRNG();
trees = (CvForestTree**)cvAlloc( sizeof(trees[0])*ntrees );
memset( trees, 0, sizeof(trees[0])*ntrees );

6
modules/ml/src/svm.cpp
Unescape View File

@ -1612,7 +1612,7 @@ bool CvSVM::train_auto( const CvMat* _train_data, const CvMat* _responses,
int block_size = 1 << 16;
double* alpha;
int i, k;
CvRNG rng = cvRNG(-1);
RNG* rng = &theRNG();
// all steps are logarithmic and must be > 1
double degree_step = 10, g_step = 10, coef_step = 10, C_step = 10, nu_step = 10, p_step = 10;
@ -1745,8 +1745,8 @@ bool CvSVM::train_auto( const CvMat* _train_data, const CvMat* _responses,
// randomly permute samples and responses
for( i = 0; i < sample_count; i++ )
{
int i1 = cvRandInt( &rng ) % sample_count;
int i2 = cvRandInt( &rng ) % sample_count;
int i1 = (*rng)(sample_count);
int i2 = (*rng)(sample_count);
const float* temp;
float t;
int y;

21
modules/ml/src/tree.cpp
Unescape View File

@ -196,7 +196,7 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
clear();
var_all = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
CV_CALL( set_params( _params ));
@ -566,7 +566,6 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
{
unsigned short* udst = 0;
int* idst = 0;
CvRNG* r = &rng;
if (is_buf_16u)
{
@ -579,8 +578,8 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
for( i = 0; i < sample_count; i++ )
{
int a = cvRandInt(r) % sample_count;
int b = cvRandInt(r) % sample_count;
int a = (*rng)(sample_count);
int b = (*rng)(sample_count);
unsigned short unsh = (unsigned short)vi;
CV_SWAP( udst[a], udst[b], unsh );
}
@ -596,8 +595,8 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
for( i = 0; i < sample_count; i++ )
{
int a = cvRandInt(r) % sample_count;
int b = cvRandInt(r) % sample_count;
int a = (*rng)(sample_count);
int b = (*rng)(sample_count);
CV_SWAP( idst[a], idst[b], vi );
}
}
@ -1134,7 +1133,7 @@ void CvDTreeTrainData::clear()
data_root = 0;
rng = cvRNG(-1);
rng = &cv::theRNG();
}
@ -2040,14 +2039,14 @@ void CvDTree::cluster_categories( const int* vectors, int n, int m,
double* buf = (double*)cvStackAlloc( (n + k)*sizeof(buf[0]) );
double *v_weights = buf, *c_weights = buf + n;
bool modified = true;
CvRNG* r = &data->rng;
RNG* r = data->rng;
// assign labels randomly
for( i = 0; i < n; i++ )
{
int sum = 0;
const int* v = vectors + i*m;
labels[i] = i < k ? i : (cvRandInt(r) % k);
labels[i] = i < k ? i : (*r)(k);
// compute weight of each vector
for( j = 0; j < m; j++ )
@ -2057,8 +2056,8 @@ void CvDTree::cluster_categories( const int* vectors, int n, int m,
for( i = 0; i < n; i++ )
{
int i1 = cvRandInt(r) % n;
int i2 = cvRandInt(r) % n;
int i1 = (*r)(n);
int i2 = (*r)(n);
CV_SWAP( labels[i1], labels[i2], j );
}

2
modules/traincascade/boost.cpp
Unescape View File

@ -218,7 +218,7 @@ void CvCascadeBoostTrainData::setData( const CvFeatureEvaluator* _featureEvaluat
have_priors = false;
is_classifier = true;
rng = cvRNG(-1);
rng = &cv::theRNG();
set_params( _params );

Write Preview
Loading…
Cancel
Save