def test_default_optimizer(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(tol=1e-4, max_iter=25)
with warnings.catch_warnings():
warnings.simplefilter("ignore", category=ConvergenceWarning)
ssvm.fit(whas500.x, whas500.y)
assert 'rbtree' == ssvm.optimizer
def test_compare_clinical_kernel(self):
x_full, y = load_whas500()
trans = ClinicalKernelTransform()
trans.fit(x_full)
kpca = KernelPCA(kernel=trans.pairwise_kernel, copy_X=True)
xt = kpca.fit_transform(self.x)
nrsvm = FastSurvivalSVM(optimizer='rbtree',
tol=1e-8,
max_iter=500,
random_state=0)
nrsvm.fit(xt, y)
rsvm = FastKernelSurvivalSVM(optimizer='rbtree',
kernel=trans.pairwise_kernel,
tol=1e-8,
max_iter=500,
random_state=0)
rsvm.fit(self.x.values, y)
pred_nrsvm = nrsvm.predict(kpca.transform(self.x))
pred_rsvm = rsvm.predict(self.x.values)
self.assertEqual(len(pred_nrsvm), len(pred_rsvm))
c1 = concordance_index_censored(y['fstat'], y['lenfol'], pred_nrsvm)
c2 = concordance_index_censored(y['fstat'], y['lenfol'], pred_rsvm)
self.assertAlmostEqual(c1[0], c2[0])
self.assertTupleEqual(c1[1:], c2[1:])
def test_compare_clinical_kernel(make_whas500):
whas500 = make_whas500(to_numeric=True)
trans = ClinicalKernelTransform()
trans.fit(whas500.x_data_frame)
kpca = KernelPCA(kernel=trans.pairwise_kernel, copy_X=True)
xt = kpca.fit_transform(whas500.x)
nrsvm = FastSurvivalSVM(optimizer='rbtree',
tol=1e-8,
max_iter=500,
random_state=0)
nrsvm.fit(xt, whas500.y)
rsvm = FastKernelSurvivalSVM(optimizer='rbtree',
kernel=trans.pairwise_kernel,
tol=1e-8,
max_iter=500,
random_state=0)
rsvm.fit(whas500.x, whas500.y)
pred_nrsvm = nrsvm.predict(kpca.transform(whas500.x))
pred_rsvm = rsvm.predict(whas500.x)
assert len(pred_nrsvm) == len(pred_rsvm)
expected_cindex = concordance_index_censored(whas500.y['fstat'],
whas500.y['lenfol'],
pred_nrsvm)
assert_cindex_almost_equal(whas500.y['fstat'], whas500.y['lenfol'],
pred_rsvm, expected_cindex)
def test_compare_builtin_kernel(make_whas500):
whas500 = make_whas500(to_numeric=True)
x = normalize(whas500.x)
rsvm = FastKernelSurvivalSVM(optimizer='rbtree',
kernel="polynomial",
gamma=0.5,
degree=2,
tol=1e-8,
max_iter=100,
random_state=0xf38)
rsvm.fit(x, whas500.y)
pred_rsvm = rsvm.predict(x)
kpca = KernelPCA(kernel="polynomial",
copy_X=True,
gamma=0.5,
degree=2,
random_state=0xf38)
xt = kpca.fit_transform(x)
nrsvm = FastSurvivalSVM(optimizer='rbtree',
tol=1e-8,
max_iter=100,
random_state=0xf38)
nrsvm.fit(xt, whas500.y)
pred_nrsvm = nrsvm.predict(xt)
assert len(pred_nrsvm) == len(pred_rsvm)
expected_cindex = concordance_index_censored(whas500.y['fstat'],
whas500.y['lenfol'],
pred_nrsvm)
assert_cindex_almost_equal(whas500.y['fstat'], whas500.y['lenfol'],
pred_rsvm, expected_cindex)
def test_unknown_optimizer(fake_data):
x, y = fake_data
ssvm = FastKernelSurvivalSVM(optimizer='random stuff')
with pytest.raises(ValueError,
match="unknown optimizer: random stuff"):
ssvm.fit(x, y)
def test_compare_rbf(self):
x, y = load_whas500()
x = encode_categorical(standardize(x))
kpca = KernelPCA(kernel="rbf")
xt = kpca.fit_transform(x)
nrsvm = FastSurvivalSVM(optimizer='rbtree',
tol=1e-8,
max_iter=1000,
random_state=0)
nrsvm.fit(xt, y)
rsvm = FastKernelSurvivalSVM(optimizer='rbtree',
kernel="rbf",
tol=1e-8,
max_iter=1000,
random_state=0)
rsvm.fit(x, y)
pred_nrsvm = nrsvm.predict(kpca.transform(x))
pred_rsvm = rsvm.predict(x)
self.assertEqual(len(pred_nrsvm), len(pred_rsvm))
c1 = concordance_index_censored(y['fstat'], y['lenfol'], pred_nrsvm)
c2 = concordance_index_censored(y['fstat'], y['lenfol'], pred_rsvm)
self.assertAlmostEqual(c1[0], c2[0])
self.assertTupleEqual(c1[1:], c2[1:])
def test_compare_builtin_kernel(self):
x = normalize(self.x)
y = self.y
rsvm = FastKernelSurvivalSVM(optimizer='rbtree',
kernel="polynomial",
gamma=0.5,
degree=2,
tol=1e-8,
max_iter=100,
random_state=0xf38)
rsvm.fit(x, y)
pred_rsvm = rsvm.predict(x)
kpca = KernelPCA(kernel="polynomial",
copy_X=True,
gamma=0.5,
degree=2,
random_state=0xf38)
xt = kpca.fit_transform(x)
nrsvm = FastSurvivalSVM(optimizer='rbtree',
tol=1e-8,
max_iter=100,
random_state=0xf38)
nrsvm.fit(xt, y)
pred_nrsvm = nrsvm.predict(xt)
self.assertEqual(len(pred_nrsvm), len(pred_rsvm))
c1 = concordance_index_censored(y['fstat'], y['lenfol'], pred_nrsvm)
c2 = concordance_index_censored(y['fstat'], y['lenfol'], pred_rsvm)
self.assertAlmostEqual(c1[0], c2[0])
self.assertTupleEqual(c1[1:], c2[1:])
def test_fit_precomputed_kernel_invalid_shape(fake_data):
x, y = fake_data
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='precomputed', random_state=0)
with pytest.raises(ValueError,
match=r"Precomputed metric requires shape \(n_queries, n_indexed\)\. "
r"Got \(100, 11\) for 100 indexed\."):
ssvm.fit(x, y)
def test_predict_precomputed_kernel_invalid_shape(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='precomputed', random_state=0)
x = numpy.dot(self.x.values, self.x.values.T)
ssvm.fit(x, self.y)
x_new = numpy.random.randn(100, 14)
self.assertRaisesRegex(ValueError, r"Precomputed metric requires shape \(n_queries, n_indexed\)\. "
r"Got \(100, 14\) for 500 indexed\.",
ssvm.predict, x_new)
def test_fit_and_predict_rbf_avltree(self):
ssvm = FastKernelSurvivalSVM(optimizer='avltree', kernel='rbf',
tol=2e-6, max_iter=75, random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(self.x.shape[0], ssvm.coef_.shape[0])
c = ssvm.score(self.x.values, self.y)
self.assertGreaterEqual(c, 0.965)
def test_fit_and_predict_linear_regression_no_intercept(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="linear",
max_iter=50, fit_intercept=False, random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(hasattr(ssvm, "intercept_"))
pred = ssvm.predict(self.x.values)
rmse = numpy.sqrt(mean_squared_error(self.y['lenfol'], pred))
self.assertAlmostEqual(rmse, 15837.658418546907, 4)
def test_fit_precomputed_kernel_not_symmetric():
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='precomputed', random_state=0)
x = numpy.random.randn(100, 100)
x[10, 12] = -1
x[12, 10] = 9
y = Surv.from_arrays(numpy.ones(100).astype(bool), numpy.ones(100))
with pytest.raises(ValueError,
match="kernel matrix is not symmetric"):
ssvm.fit(x, y)
def test_fit_and_predict_rbf_avltree(self):
ssvm = FastKernelSurvivalSVM(optimizer="avltree",
kernel='rbf',
random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(self.x.shape[0], ssvm.coef_.shape[0])
c = ssvm.score(self.x.values, self.y)
self.assertLessEqual(abs(0.92460312179802795 - c), 1e-3)
def test_fit_and_predict_hybrid_rbf(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.5, kernel="rbf",
max_iter=50, fit_intercept=True, random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertLessEqual(abs(5.0289145697617164 - ssvm.intercept_), 0.04)
pred = ssvm.predict(self.x.values)
rmse = numpy.sqrt(mean_squared_error(self.y['lenfol'], pred))
self.assertLessEqual(abs(880.20361811281487 - rmse), 75)
def test_fit_and_predict_regression_rbf(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="rbf",
tol=1e-6, max_iter=50, fit_intercept=True, random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertAlmostEqual(ssvm.intercept_, 4.9267218894089533)
pred = ssvm.predict(self.x.values)
rmse = numpy.sqrt(mean_squared_error(self.y['lenfol'], pred))
self.assertAlmostEqual(rmse, 783.525277, 6)
def test_fit_and_predict_linear(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='linear', random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(self.x.shape[0], ssvm.coef_.shape[0])
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
c = ssvm.score(self.x.values[i], self.y[i])
self.assertAlmostEqual(c, 0.76923445664157997, 6)
def test_predict_precomputed_kernel_invalid_shape(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='precomputed', random_state=0)
x = numpy.dot(whas500.x, whas500.x.T)
ssvm.fit(x, whas500.y)
x_new = numpy.random.randn(100, 14)
with pytest.raises(ValueError,
match=r"Precomputed metric requires shape \(n_queries, n_indexed\)\. "
r"Got \(100, 14\) for 500 indexed\."):
ssvm.predict(x_new)
def test_fit_and_predict_linear_regression_no_intercept(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="linear",
max_iter=50, fit_intercept=False, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not hasattr(ssvm, "intercept_")
pred = ssvm.predict(whas500.x)
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'], pred))
assert round(abs(rmse - 15837.658418546907), 4) == 0
def test_fit_and_predict_rbf_rbtree(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree",
kernel='rbf',
random_state=0)
ssvm.fit(self.x.values, self.y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(self.x.shape[0], ssvm.coef_.shape[0])
c = ssvm.score(self.x.values, self.y)
self.assertAlmostEqual(0.92230102862313534, c, 3)
def test_fit_and_predict_rbf(make_whas500, optimizer):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer=optimizer, kernel='rbf',
tol=2e-6, max_iter=75, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert whas500.x.shape[0] == ssvm.coef_.shape[0]
c = ssvm.score(whas500.x, whas500.y)
assert c >= 0.965
def test_fit_and_predict_hybrid_rbf(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.5, kernel="rbf",
max_iter=50, fit_intercept=True, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert abs(5.0289145697617164 - ssvm.intercept_) <= 0.04
pred = ssvm.predict(whas500.x)
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'], pred))
assert abs(880.20361811281487 - rmse) <= 75
def test_fit_and_predict_regression_rbf(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="rbf",
tol=1e-6, max_iter=50, fit_intercept=True, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert round(abs(ssvm.intercept_ - 4.9267218894089533), 7) == 0
pred = ssvm.predict(whas500.x)
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'], pred))
assert round(abs(rmse - 783.525277), 6) == 0
def test_fit_and_predict_linear(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel='linear', random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert whas500.x.shape[0] == ssvm.coef_.shape[0]
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
c = ssvm.score(whas500.x[i], whas500.y[i])
assert round(abs(c - 0.76923445664157997), 6) == 0
def test_fit_and_predict_linear_regression_precomputed(self):
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="precomputed",
max_iter=50, fit_intercept=True, random_state=0)
x = numpy.dot(self.x.values, self.x.values.T)
ssvm.fit(x, self.y)
self.assertTrue(ssvm._pairwise)
self.assertAlmostEqual(ssvm.intercept_, 6.3979746625712295, 5)
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
pred = ssvm.predict(x[i])
rmse = numpy.sqrt(mean_squared_error(self.y['lenfol'][i], pred))
self.assertLessEqual(abs(1339.3006854574726 - rmse), 0.25)
def test_fit_and_predict_clinical_kernel(make_whas500):
whas500 = make_whas500(to_numeric=True)
trans = ClinicalKernelTransform()
trans.fit(whas500.x_data_frame)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel=trans.pairwise_kernel,
tol=7e-7, max_iter=100, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert whas500.x.shape[0] == ssvm.coef_.shape[0]
c = ssvm.score(whas500.x, whas500.y)
assert c >= 0.854
def test_fit_and_predict_linear_regression_precomputed(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="precomputed",
max_iter=50, fit_intercept=True, random_state=0)
x = numpy.dot(whas500.x, whas500.x.T)
ssvm.fit(x, whas500.y)
assert ssvm._pairwise
assert round(abs(ssvm.intercept_ - 6.3979746625712295), 5) == 0
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
pred = ssvm.predict(x[i])
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'][i], pred))
assert rmse <= 1339.3006854574726 + 0.293
def test_fit_and_predict_clinical_kernel(self):
x_full, y = load_whas500()
trans = ClinicalKernelTransform()
trans.fit(x_full)
x = self.x
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", kernel=trans.pairwise_kernel,
tol=7e-7, max_iter=100, random_state=0)
ssvm.fit(x.values, y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(x.shape[0], ssvm.coef_.shape[0])
c = ssvm.score(x.values, y)
self.assertGreaterEqual(c, 0.854)
def test_fit_and_predict_linear_regression(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree", rank_ratio=0.0, kernel="linear",
max_iter=50, tol=1e-8, fit_intercept=True, random_state=0)
ssvm.fit(whas500.x, whas500.y)
assert not ssvm._pairwise
assert round(abs(ssvm.intercept_ - 6.416017539824949), 5) == 0
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
pred = ssvm.predict(whas500.x[i])
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'][i], pred))
assert rmse <= 1342.274550652291 + 0.293
c = ssvm.score(whas500.x[i], whas500.y[i])
assert round(abs(c - 0.7630027323714108), 6) == 0
def test_fit_and_predict_clinical_kernel(self):
x_full, y = load_whas500()
trans = ClinicalKernelTransform()
trans.fit(x_full)
x = encode_categorical(standardize(x_full))
ssvm = FastKernelSurvivalSVM(optimizer="rbtree",
kernel=trans.pairwise_kernel,
max_iter=100,
random_state=0)
ssvm.fit(x.values, y)
self.assertFalse(ssvm._pairwise)
self.assertEquals(x.shape[0], ssvm.coef_.shape[0])
c = ssvm.score(x.values, y)
self.assertLessEqual(abs(0.83699051218246412 - c), 1e-3)
def test_fit_and_predict_linear_regression_precomputed(make_whas500):
whas500 = make_whas500(to_numeric=True)
ssvm = FastKernelSurvivalSVM(optimizer="rbtree",
rank_ratio=0.0,
kernel="precomputed",
max_iter=50,
tol=1e-8,
fit_intercept=True,
random_state=0)
x = numpy.dot(whas500.x, whas500.x.T)
ssvm.fit(x, whas500.y)
assert ssvm._get_tags()["pairwise"]
assert round(abs(ssvm.intercept_ - 6.416017539824949), 5) == 0
i = numpy.arange(250)
numpy.random.RandomState(0).shuffle(i)
pred = ssvm.predict(x[i])
rmse = numpy.sqrt(mean_squared_error(whas500.y['lenfol'][i], pred))
assert rmse <= 1342.274550652291 + 0.293