MCPcopy Create free account
hub / github.com/davisking/dlib / cross_validate_trainer_impl

Function cross_validate_trainer_impl

dlib/svm/svm.h:407–534  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

405 typename in_scalar_vector_type
406 >
407 const matrix<double, 1, 2, typename trainer_type::mem_manager_type>
408 cross_validate_trainer_impl (
409 const trainer_type& trainer,
410 const in_sample_vector_type& x,
411 const in_scalar_vector_type& y,
412 const long folds
413 )
414 {
415 typedef typename in_scalar_vector_type::value_type scalar_type;
416 typedef typename trainer_type::mem_manager_type mem_manager_type;
417 typedef matrix<scalar_type,0,1,mem_manager_type> scalar_vector_type;
418
419 // make sure requires clause is not broken
420 DLIB_CASSERT(is_binary_classification_problem(x,y) == true &&
421 1 < folds && folds <= std::min(sum(y>0),sum(y<0)),
422 "\tmatrix cross_validate_trainer()"
423 << "\n\t invalid inputs were given to this function"
424 << "\n\t std::min(sum(y>0),sum(y<0)): " << std::min(sum(y>0),sum(y<0))
425 << "\n\t folds: " << folds
426 << "\n\t is_binary_classification_problem(x,y): " << ((is_binary_classification_problem(x,y))? "true":"false")
427 );
428
429
430 // count the number of positive and negative examples
431 long num_pos = 0;
432 long num_neg = 0;
433 for (long r = 0; r < y.nr(); ++r)
434 {
435 if (y(r) == +1.0)
436 ++num_pos;
437 else
438 ++num_neg;
439 }
440
441 // figure out how many positive and negative examples we will have in each fold
442 const long num_pos_test_samples = num_pos/folds;
443 const long num_pos_train_samples = num_pos - num_pos_test_samples;
444 const long num_neg_test_samples = num_neg/folds;
445 const long num_neg_train_samples = num_neg - num_neg_test_samples;
446
447
448 matrix<long,0,1> x_test, x_train;
449 scalar_vector_type y_test, y_train;
450 x_test.set_size (num_pos_test_samples + num_neg_test_samples);
451 y_test.set_size (num_pos_test_samples + num_neg_test_samples);
452 x_train.set_size(num_pos_train_samples + num_neg_train_samples);
453 y_train.set_size(num_pos_train_samples + num_neg_train_samples);
454
455 long pos_idx = 0;
456 long neg_idx = 0;
457
458 matrix<double, 1, 2, mem_manager_type> res;
459 set_all_elements(res,0);
460
461 for (long i = 0; i < folds; ++i)
462 {
463 long cur = 0;
464

Callers 1

cross_validate_trainerFunction · 0.85

Calls 10

yFunction · 0.85
set_all_elementsFunction · 0.85
minFunction · 0.70
sumFunction · 0.50
rowmFunction · 0.50
nrMethod · 0.45
set_sizeMethod · 0.45
trainMethod · 0.45

Tested by

no test coverage detected