def test_fit_rbf_binary(): for selection in ("permute", "active", "loss"): clf = LaSVM(random_state=0, max_iter=2, kernel="rbf", selection=selection) clf.fit(bin_dense, bin_target) acc = clf.score(bin_dense, bin_target) assert_almost_equal(acc, 1.0)
def test_fit_linear_binary(): for selection, exp in (("permute", 1.0), ("active", 1.0), ("loss", 1.0)): clf = LaSVM(random_state=0, max_iter=2, kernel="linear", selection=selection) clf.fit(bin_dense, bin_target) acc = clf.score(bin_dense, bin_target) assert_almost_equal(acc, exp)
def test_warm_start(): for selection in ("permute", "active", "loss"): clf = LaSVM(random_state=0, max_iter=2, kernel="rbf", warm_start=True, selection=selection) clf.C = 0.5 clf.fit(bin_dense, bin_target) acc = clf.score(bin_dense, bin_target) assert_almost_equal(acc, 1.0, 1) clf.C = 0.6 clf.fit(bin_dense, bin_target) acc = clf.score(bin_dense, bin_target) assert_almost_equal(acc, 1.0)
def test_sv_upper_bound(): clf = LaSVM(random_state=0, max_iter=2, kernel="rbf", finish_step=True, termination="n_sv", sv_upper_bound=30) clf.fit(bin_dense, bin_target) n_sv = np.sum(clf.dual_coef_ != 0) assert_equal(n_sv, 30)
def test_fit_rbf_multi(): clf = LaSVM(kernel="rbf", gamma=0.1, random_state=0) clf.fit(mult_dense, mult_target) y_pred = clf.predict(mult_dense) acc = np.mean(y_pred == mult_target) assert_almost_equal(acc, 1.0)
def test_n_components(): clf = LaSVM(random_state=0, max_iter=2, kernel="rbf", finish_step=True, termination="n_components", n_components=30) clf.fit(bin_dense, bin_target) n_sv = np.sum(clf.coef_ != 0) assert_equal(n_sv, 30)