| 486 | // ---------------------------------------------------------------------------------------- |
| 487 | |
| 488 | void unittest_binary_classification ( |
| 489 | ) |
| 490 | /*! |
| 491 | ensures |
| 492 | - runs tests on the svm stuff compliance with the specs |
| 493 | !*/ |
| 494 | { |
| 495 | dlog << LINFO << " begin unittest_binary_classification()"; |
| 496 | print_spinner(); |
| 497 | |
| 498 | |
| 499 | typedef double scalar_type; |
| 500 | typedef matrix<scalar_type,2,1> sample_type; |
| 501 | |
| 502 | std::vector<sample_type> x; |
| 503 | std::vector<matrix<double,0,1> > x_linearized; |
| 504 | std::vector<scalar_type> y; |
| 505 | |
| 506 | get_checkerboard_problem(x,y, 300, 2); |
| 507 | const scalar_type gamma = 1; |
| 508 | |
| 509 | typedef radial_basis_kernel<sample_type> kernel_type; |
| 510 | |
| 511 | rbf_network_trainer<kernel_type> rbf_trainer; |
| 512 | rbf_trainer.set_kernel(kernel_type(gamma)); |
| 513 | rbf_trainer.set_num_centers(100); |
| 514 | |
| 515 | rvm_trainer<kernel_type> rvm_trainer; |
| 516 | rvm_trainer.set_kernel(kernel_type(gamma)); |
| 517 | |
| 518 | krr_trainer<kernel_type> krr_trainer; |
| 519 | krr_trainer.use_classification_loss_for_loo_cv(); |
| 520 | krr_trainer.set_kernel(kernel_type(gamma)); |
| 521 | |
| 522 | svm_pegasos<kernel_type> pegasos_trainer; |
| 523 | pegasos_trainer.set_kernel(kernel_type(gamma)); |
| 524 | pegasos_trainer.set_lambda(0.00001); |
| 525 | |
| 526 | |
| 527 | svm_c_ekm_trainer<kernel_type> ocas_ekm_trainer; |
| 528 | ocas_ekm_trainer.set_kernel(kernel_type(gamma)); |
| 529 | ocas_ekm_trainer.set_c(100000); |
| 530 | |
| 531 | svm_nu_trainer<kernel_type> trainer; |
| 532 | trainer.set_kernel(kernel_type(gamma)); |
| 533 | trainer.set_nu(0.05); |
| 534 | |
| 535 | svm_c_trainer<kernel_type> c_trainer; |
| 536 | c_trainer.set_kernel(kernel_type(gamma)); |
| 537 | c_trainer.set_c(100); |
| 538 | |
| 539 | svm_c_linear_trainer<linear_kernel<matrix<double,0,1> > > lin_trainer; |
| 540 | lin_trainer.set_c(100000); |
| 541 | // use an ekm to linearize this dataset so we can use it with the lin_trainer |
| 542 | empirical_kernel_map<kernel_type> ekm; |
| 543 | ekm.load(kernel_type(gamma), x); |
| 544 | for (unsigned long i = 0; i < x.size(); ++i) |
| 545 | x_linearized.push_back(ekm.project(x[i])); |
no test coverage detected