third_party/opencv
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #720 from taka-no-me:drop_sort

Browse Source
This commit is contained in:
Andrey Kamaev
2013-04-01 15:14:45 +04:00
committed by OpenCV Buildbot
parent f77a375354 855acc9748
commit db45e04d58
20 changed files with 144 additions and 577 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/ml/src/boost.cpp
+9 -9
View File
@@ -351,9 +351,12 @@ CvBoostTree::find_split_ord_class( CvDTreeNode* node, int vi, float init_quality
return split;
}
#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
static CV_IMPLEMENT_QSORT_EX( icvSortDblPtr, double*, CV_CMP_NUM_PTR, int )
template<typename T>
class LessThanPtr
{
public:
bool operator()(T* a, T* b) const { return *a < *b; }
};
CvDTreeSplit*
CvBoostTree::find_split_cat_class( CvDTreeNode* node, int vi, float init_quality, CvDTreeSplit* _split, uchar* _ext_buf )
@@ -412,7 +415,7 @@ CvBoostTree::find_split_cat_class( CvDTreeNode* node, int vi, float init_quality
// sort rows of c_jk by increasing c_j,1
// (i.e. by the weight of samples in j-th category that belong to class 1)
icvSortDblPtr( dbl_ptr, mi, 0 );
std::sort(dbl_ptr, dbl_ptr + mi, LessThanPtr<double>());
for( subset_i = 0; subset_i < mi-1; subset_i++ )
{
@@ -594,7 +597,7 @@ CvBoostTree::find_split_cat_reg( CvDTreeNode* node, int vi, float init_quality,
sum_ptr[i] = sum + i;
}
icvSortDblPtr( sum_ptr, mi, 0 );
std::sort(sum_ptr, sum_ptr + mi, LessThanPtr<double>());
// revert back to unnormalized sums
// (there should be a very little loss in accuracy)
@@ -1421,9 +1424,6 @@ CvBoost::update_weights( CvBoostTree* tree )
}
static CV_IMPLEMENT_QSORT_EX( icvSort_64f, double, CV_LT, int )
void
CvBoost::trim_weights()
{
@@ -1440,7 +1440,7 @@ CvBoost::trim_weights()
// use weak_eval as temporary buffer for sorted weights
cvCopy( weights, weak_eval );
icvSort_64f( weak_eval->data.db, count, 0 );
std::sort(weak_eval->data.db, weak_eval->data.db + count);
// as weight trimming occurs immediately after updating the weights,
// where they are renormalized, we assume that the weight sum = 1.
modules/ml/src/ertrees.cpp
+13 -8
View File
@@ -44,13 +44,18 @@ static const float ord_nan = FLT_MAX*0.5f;
static const int min_block_size = 1 << 16;
static const int block_size_delta = 1 << 10;
#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
static CV_IMPLEMENT_QSORT_EX( icvSortIntPtr, int*, CV_CMP_NUM_PTR, int )
template<typename T>
class LessThanPtr
{
public:
bool operator()(T* a, T* b) const { return *a < *b; }
};
#define CV_CMP_PAIRS(a,b) (*((a).i) < *((b).i))
static CV_IMPLEMENT_QSORT_EX( icvSortPairs, CvPair16u32s, CV_CMP_PAIRS, int )
///
class LessThanPairs
{
public:
bool operator()(const CvPair16u32s& a, const CvPair16u32s& b) const { return *a.i < *b.i; }
};
void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
const CvMat* _responses, const CvMat* _var_idx, const CvMat* _sample_idx,
@@ -353,7 +358,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
if (is_buf_16u)
{
icvSortPairs( pair16u32s_ptr, sample_count, 0 );
std::sort(pair16u32s_ptr, pair16u32s_ptr + sample_count, LessThanPairs());
// count the categories
for( i = 1; i < num_valid; i++ )
if (*pair16u32s_ptr[i].i != *pair16u32s_ptr[i-1].i)
@@ -361,7 +366,7 @@ void CvERTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
}
else
{
icvSortIntPtr( int_ptr, sample_count, 0 );
std::sort(int_ptr, int_ptr + sample_count, LessThanPtr<int>());
// count the categories
for( i = 1; i < num_valid; i++ )
c_count += *int_ptr[i] != *int_ptr[i-1];
modules/ml/src/gbt.cpp
+4 -7
View File
@@ -5,9 +5,6 @@
#define pCvSeq CvSeq*
#define pCvDTreeNode CvDTreeNode*
#define CV_CMP_FLOAT(a,b) ((a) < (b))
static CV_IMPLEMENT_QSORT_EX( icvSortFloat, float, CV_CMP_FLOAT, float)
//===========================================================================
//----------------------------- CvGBTreesParams -----------------------------
//===========================================================================
@@ -285,7 +282,7 @@ CvGBTrees::train( const CvMat* _train_data, int _tflag,
} break;
default: CV_Error(CV_StsUnmatchedFormats, "_sample_idx should be a 32sC1, 8sC1 or 8uC1 vector.");
}
icvSortFloat(sample_idx->data.fl, sample_idx_len, 0);
std::sort(sample_idx->data.fl, sample_idx->data.fl + sample_idx_len);
}
else
{
@@ -470,7 +467,7 @@ void CvGBTrees::find_gradient(const int k)
int idx = *(sample_data + subsample_data[i]*s_step);
residuals[i] = fabs(resp_data[idx] - current_data[idx]);
}
icvSortFloat(residuals, n, 0.0f);
std::sort(residuals, residuals + n);
delta = residuals[int(ceil(n*alpha))];
@@ -693,7 +690,7 @@ float CvGBTrees::find_optimal_value( const CvMat* _Idx )
float* residuals = new float[n];
for (int i=0; i<n; ++i, ++idx)
residuals[i] = (resp_data[*idx] - cur_data[*idx]);
icvSortFloat(residuals, n, 0.0f);
std::sort(residuals, residuals + n);
if (n % 2)
gamma = residuals[n/2];
else gamma = (residuals[n/2-1] + residuals[n/2]) / 2.0f;
@@ -705,7 +702,7 @@ float CvGBTrees::find_optimal_value( const CvMat* _Idx )
float* residuals = new float[n];
for (int i=0; i<n; ++i, ++idx)
residuals[i] = (resp_data[*idx] - cur_data[*idx]);
icvSortFloat(residuals, n, 0.0f);
std::sort(residuals, residuals + n);
int n_half = n >> 1;
float r_median = (n == n_half<<1) ?
modules/ml/src/tree.cpp
+25 -14
View File
@@ -120,16 +120,27 @@ bool CvDTreeTrainData::set_params( const CvDTreeParams& _params )
return ok;
}
#define CV_CMP_NUM_PTR(a,b) (*(a) < *(b))
static CV_IMPLEMENT_QSORT_EX( icvSortIntPtr, int*, CV_CMP_NUM_PTR, int )
static CV_IMPLEMENT_QSORT_EX( icvSortDblPtr, double*, CV_CMP_NUM_PTR, int )
template<typename T>
class LessThanPtr
{
public:
bool operator()(T* a, T* b) const { return *a < *b; }
};
#define CV_CMP_NUM_IDX(i,j) (aux[i] < aux[j])
static CV_IMPLEMENT_QSORT_EX( icvSortIntAux, int, CV_CMP_NUM_IDX, const float* )
static CV_IMPLEMENT_QSORT_EX( icvSortUShAux, unsigned short, CV_CMP_NUM_IDX, const float* )
template<typename T, typename Idx>
class LessThanIdx
{
public:
LessThanIdx( const T* _arr ) : arr(_arr) {}
bool operator()(Idx a, Idx b) const { return arr[a] < arr[b]; }
const T* arr;
};
#define CV_CMP_PAIRS(a,b) (*((a).i) < *((b).i))
static CV_IMPLEMENT_QSORT_EX( icvSortPairs, CvPair16u32s, CV_CMP_PAIRS, int )
class LessThanPairs
{
public:
bool operator()(const CvPair16u32s& a, const CvPair16u32s& b) const { return *a.i < *b.i; }
};
void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
const CvMat* _responses, const CvMat* _var_idx, const CvMat* _sample_idx,
@@ -461,7 +472,7 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
c_count = num_valid > 0;
if (is_buf_16u)
{
icvSortPairs( pair16u32s_ptr, sample_count, 0 );
std::sort(pair16u32s_ptr, pair16u32s_ptr + sample_count, LessThanPairs());
// count the categories
for( i = 1; i < num_valid; i++ )
if (*pair16u32s_ptr[i].i != *pair16u32s_ptr[i-1].i)
@@ -469,7 +480,7 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
}
else
{
icvSortIntPtr( int_ptr, sample_count, 0 );
std::sort(int_ptr, int_ptr + sample_count, LessThanPtr<int>());
// count the categories
for( i = 1; i < num_valid; i++ )
c_count += *int_ptr[i] != *int_ptr[i-1];
@@ -561,9 +572,9 @@ void CvDTreeTrainData::set_data( const CvMat* _train_data, int _tflag,
}
if (is_buf_16u)
icvSortUShAux( udst, sample_count, _fdst);
std::sort(udst, udst + sample_count, LessThanIdx<float, unsigned short>(_fdst));
else
icvSortIntAux( idst, sample_count, _fdst );
std::sort(idst, idst + sample_count, LessThanIdx<float, int>(_fdst));
}
if( vi < var_count )
@@ -2239,7 +2250,7 @@ CvDTreeSplit* CvDTree::find_split_cat_class( CvDTreeNode* node, int vi, float in
int_ptr = (int**)(c_weights + _mi);
for( j = 0; j < mi; j++ )
int_ptr[j] = cjk + j*2 + 1;
icvSortIntPtr( int_ptr, mi, 0 );
std::sort(int_ptr, int_ptr + mi, LessThanPtr<int>());
subset_i = 0;
subset_n = mi;
}
@@ -2466,7 +2477,7 @@ CvDTreeSplit* CvDTree::find_split_cat_reg( CvDTreeNode* node, int vi, float init
sum_ptr[i] = sum + i;
}
icvSortDblPtr( sum_ptr, mi, 0 );
std::sort(sum_ptr, sum_ptr + mi, LessThanPtr<double>());
// revert back to unnormalized sums
// (there should be a very little loss of accuracy)