def test_predict():
iris = datasets.load_iris()
X = iris.data
y = iris.target
reg = SVR(kernel='linear', C=0.1).fit(X, y)
linear = np.dot(X, reg.support_vectors_.T)
dec = np.dot(linear, reg.dual_coef_.T) + reg.intercept_
assert_array_almost_equal(dec.ravel(), reg.predict(X).ravel())
reg = SVR(kernel='rbf', gamma=1).fit(X, y)
rbfs = rbf_kernel(X, reg.support_vectors_, gamma=reg.gamma)
dec = np.dot(rbfs, reg.dual_coef_.T) + reg.intercept_
assert_array_almost_equal(dec.ravel(), reg.predict(X).ravel())
def test_sided_sample_weight():
clf = SVR(C=1e-2, kernel='linear')
X = [[-2, 0], [-1, -1], [0, -2], [0, 2], [1, 1], [2, 0]]
Y = [1, 1, 1, 2, 2, 2]
sample_weight = [10., .1, .1, .1, .1, 10]
clf.fit(X, Y, sample_weight=sample_weight)
y_pred = clf.predict([[-1., 1.]])
assert y_pred < 1.5
sample_weight = [1., .1, 10., 10., .1, .1]
clf.fit(X, Y, sample_weight=sample_weight)
y_pred = clf.predict([[-1., 1.]])
assert y_pred > 1.5
sample_weight = [1] * 6
clf.fit(X, Y, sample_weight=sample_weight)
y_pred = clf.predict([[-1., 1.]])
assert y_pred == pytest.approx(1.5)
def test_pickle():
diabetes = datasets.load_diabetes()
clf = SVR(kernel='rbf', C=10.)
clf.fit(diabetes.data, diabetes.target)
expected = clf.predict(diabetes.data)
import pickle
dump = pickle.dumps(clf)
clf2 = pickle.loads(dump)
assert type(clf2) == clf.__class__
result = clf2.predict(diabetes.data)
assert_array_equal(expected, result)
def test_input_format_for_diabetes():
diabetes = datasets.load_diabetes()
c_contiguous_numpy = np.asanyarray(diabetes.data, dtype='float', order='C')
assert c_contiguous_numpy.flags.c_contiguous
assert not c_contiguous_numpy.flags.f_contiguous
assert not c_contiguous_numpy.flags.fnc
clf = SVR(kernel='linear', C=10.)
clf.fit(c_contiguous_numpy, diabetes.target)
dual_c_contiguous_numpy = clf.dual_coef_
res_c_contiguous_numpy = clf.predict(c_contiguous_numpy)
f_contiguous_numpy = np.asanyarray(diabetes.data, dtype='float', order='F')
assert not f_contiguous_numpy.flags.c_contiguous
assert f_contiguous_numpy.flags.f_contiguous
assert f_contiguous_numpy.flags.fnc
clf = SVR(kernel='linear', C=10.)
clf.fit(f_contiguous_numpy, diabetes.target)
dual_f_contiguous_numpy = clf.dual_coef_
res_f_contiguous_numpy = clf.predict(f_contiguous_numpy)
assert_allclose(dual_c_contiguous_numpy, dual_f_contiguous_numpy)
assert_allclose(res_c_contiguous_numpy, res_f_contiguous_numpy)
