From ffe79c9044059fed487eb05d3e3c51d4058a3383 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sebastian=20P=C3=B6lsterl?= <sebp@k-d-w.org>
Date: Sat, 27 Jun 2020 17:44:20 +0200
Subject: [PATCH] Throw exception when trying to fit model to data with
uncomparable pairs
---
sksurv/svm/minlip.py | 4 ++++
sksurv/svm/naive_survival_svm.py | 3 +++
sksurv/svm/survival_svm.py | 35 +++++++++++++++++++++++++-------
tests/conftest.py | 9 ++++++++
tests/test_minlip.py | 16 +++++++++++++--
tests/test_survival_svm.py | 21 +++++++++++++++++++
6 files changed, 79 insertions(+), 9 deletions(-)
diff --git a/sksurv/svm/minlip.py b/sksurv/svm/minlip.py
index 2722a98a..53e177d7 100644
--- a/sksurv/svm/minlip.py
+++ b/sksurv/svm/minlip.py
@@ -6,6 +6,7 @@
import warnings
from ..base import SurvivalAnalysisMixin
+from ..exceptions import NoComparablePairException
from ..util import check_arrays_survival
from ._minlip import create_difference_matrix
@@ -135,6 +136,9 @@ def _get_kernel(self, X, Y=None):
def _fit(self, x, event, time):
D = create_difference_matrix(event.astype(numpy.uint8), time, kind=self.pairs)
+ if D.shape[0] == 0:
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
+
K = self._get_kernel(x)
if self.solver == "cvxpy":
diff --git a/sksurv/svm/naive_survival_svm.py b/sksurv/svm/naive_survival_svm.py
index 6439fad0..d6a573f7 100644
--- a/sksurv/svm/naive_survival_svm.py
+++ b/sksurv/svm/naive_survival_svm.py
@@ -18,6 +18,7 @@
from sklearn.utils import check_random_state
from ..base import SurvivalAnalysisMixin
+from ..exceptions import NoComparablePairException
from ..util import check_arrays_survival
@@ -154,6 +155,8 @@ def fit(self, X, y, sample_weight=None):
random_state = check_random_state(self.random_state)
x_pairs, y_pairs = self._get_survival_pairs(X, y, random_state)
+ if x_pairs.shape[0] == 0:
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
self.C = self.alpha
return super().fit(x_pairs, y_pairs, sample_weight=sample_weight)
diff --git a/sksurv/svm/survival_svm.py b/sksurv/svm/survival_svm.py
index 083f4d4a..2b87ac94 100644
--- a/sksurv/svm/survival_svm.py
+++ b/sksurv/svm/survival_svm.py
@@ -11,22 +11,22 @@
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
from abc import ABCMeta, abstractmethod
+import warnings
+
+import numpy
+import numexpr
+from scipy.optimize import minimize
from sklearn.base import BaseEstimator
from sklearn.exceptions import ConvergenceWarning
from sklearn.metrics.pairwise import pairwise_kernels
from sklearn.utils import check_X_y, check_array, check_consistent_length, check_random_state
from sklearn.utils.extmath import safe_sparse_dot, squared_norm
-from scipy.optimize import minimize
-
-import numpy
-import numexpr
-import warnings
-
-from ._prsvm import survival_constraints_simple, survival_constraints_with_support_vectors
from ..base import SurvivalAnalysisMixin
from ..bintrees import AVLTree, RBTree
+from ..exceptions import NoComparablePairException
from ..util import check_arrays_survival
+from ._prsvm import survival_constraints_simple, survival_constraints_with_support_vectors
class Counter(object, metaclass=ABCMeta):
@@ -263,6 +263,9 @@ def __init__(self, x, y, alpha, rank_ratio, timeit=False):
self.data_x = x
self.constraints = survival_constraints_simple(numpy.asarray(y, dtype=numpy.uint8))
+ if self.constraints.shape[0] == 0:
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
+
self.L = numpy.ones(self.constraints.shape[0])
@property
@@ -303,6 +306,10 @@ def __init__(self, x, y, alpha, rank_ratio, timeit=False):
self.data_y = numpy.asarray(y, dtype=numpy.uint8)
self._constraints = lambda w: survival_constraints_with_support_vectors(self.data_y, w)
+ Aw = self._constraints(numpy.zeros(x.shape[1]))
+ if Aw.shape[0] == 0:
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
+
@property
def n_coefficients(self):
return self.data_x.shape[1]
@@ -372,8 +379,15 @@ def n_coefficients(self):
def _init_coefficients(self):
w = super()._init_coefficients()
+ n = w.shape[0]
if self._fit_intercept:
w[0] = self._counter.time.mean()
+ n -= 1
+
+ l_plus, _, l_minus, _ = self._counter.calculate(numpy.zeros(n))
+ if numpy.all(l_plus == 0) and numpy.all(l_minus == 0):
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
+
return w
def _split_coefficents(self, w):
@@ -498,8 +512,15 @@ def n_coefficients(self):
def _init_coefficients(self):
w = super()._init_coefficients()
+ n = w.shape[0]
if self._fit_intercept:
w[0] = self._counter.time.mean()
+ n -= 1
+
+ l_plus, _, l_minus, _ = self._counter.calculate(numpy.zeros(n))
+ if numpy.all(l_plus == 0) and numpy.all(l_minus == 0):
+ raise NoComparablePairException("Data has no comparable pairs, cannot fit model.")
+
return w
def _split_coefficents(self, w):
diff --git a/tests/conftest.py b/tests/conftest.py
index 7270fe5b..a21e35dd 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -67,6 +67,15 @@ def whas500_sparse_data():
return SparseDataSet(x_dense=x_dense, x_sparse=x_sparse, y=y)
+@pytest.fixture
+def whas500_uncomparable(make_whas500):
+ whas500 = make_whas500(to_numeric=True)
+ i = numpy.argmax(whas500.y["lenfol"])
+ whas500.y["fstat"][:] = False
+ whas500.y["fstat"][i] = True
+ return whas500
+
+
@pytest.fixture
def rossi():
"""Load rossi.csv"""
diff --git a/tests/test_minlip.py b/tests/test_minlip.py
index dd80c0ec..31d11ab8 100644
--- a/tests/test_minlip.py
+++ b/tests/test_minlip.py
@@ -1,14 +1,16 @@
+from itertools import product
+
import numpy
from numpy.testing import assert_array_almost_equal, assert_array_equal
import pytest
-
from sklearn.exceptions import ConvergenceWarning
from sklearn.preprocessing import scale
-from sksurv.svm.minlip import MinlipSurvivalAnalysis, HingeLossSurvivalSVM
from sksurv.datasets import load_gbsg2
+from sksurv.exceptions import NoComparablePairException
from sksurv.column import encode_categorical
from sksurv.svm._minlip import create_difference_matrix
+from sksurv.svm.minlip import MinlipSurvivalAnalysis, HingeLossSurvivalSVM
from sksurv.testing import assert_cindex_almost_equal
from sksurv.util import Surv
@@ -693,3 +695,13 @@ def test_max_iter(gbsg2):
with pytest.warns(ConvergenceWarning,
match=r"cvxopt solver did not converge: unknown \(duality gap = [.0-9]+\)"):
m.fit(x, y)
+
+
+@pytest.mark.parametrize(["model_cls", "solver", "pairs"],
+ list(product((MinlipSurvivalAnalysis, HingeLossSurvivalSVM),
+ ("cvxpy", "cvxopt", "osqp"),
+ ("all", "nearest", "next"))))
+def test_fit_uncomparable(whas500_uncomparable, model_cls, solver, pairs):
+ ssvm = model_cls(solver=solver, pairs=pairs)
+ with pytest.raises(NoComparablePairException):
+ ssvm.fit(whas500_uncomparable.x, whas500_uncomparable.y)
diff --git a/tests/test_survival_svm.py b/tests/test_survival_svm.py
index 40fbccab..49f02625 100644
--- a/tests/test_survival_svm.py
+++ b/tests/test_survival_svm.py
@@ -13,6 +13,7 @@
from sksurv.bintrees import AVLTree, RBTree
from sksurv.column import encode_categorical
from sksurv.datasets import load_whas500, get_x_y
+from sksurv.exceptions import NoComparablePairException
from sksurv.io import loadarff
from sksurv.kernels import ClinicalKernelTransform
from sksurv.metrics import concordance_index_censored
@@ -177,6 +178,13 @@ def test_ranking_with_fit_intercept():
match="fit_intercept=True is only meaningful if rank_ratio < 1.0"):
ssvm.fit(x, y)
+ @staticmethod
+ @pytest.mark.parametrize("optimizer", ("simple", "avltree", "direct-count", "PRSVM", "rbtree"))
+ def test_fit_uncomparable(whas500_uncomparable, optimizer):
+ ssvm = FastSurvivalSVM(optimizer=optimizer)
+ with pytest.raises(NoComparablePairException):
+ ssvm.fit(whas500_uncomparable.x, whas500_uncomparable.y)
+
@staticmethod
def test_survial_constraints_no_ties():
y = numpy.array([True, True, False, True, False, False, False, False])
@@ -629,6 +637,13 @@ def test_predict_precomputed_kernel_invalid_shape(make_whas500):
r"Got \(100, 14\) for 500 indexed\."):
ssvm.predict(x_new)
+ @staticmethod
+ @pytest.mark.parametrize("optimizer", ("avltree", "rbtree"))
+ def test_fit_uncomparable(whas500_uncomparable, optimizer):
+ ssvm = FastKernelSurvivalSVM(optimizer=optimizer)
+ with pytest.raises(NoComparablePairException):
+ ssvm.fit(whas500_uncomparable.x, whas500_uncomparable.y)
+
@pytest.fixture(params=[
SurvivalCounter,
@@ -766,3 +781,9 @@ def test_fit_with_ties(whas500_with_ties):
cindex = nrsvm.score(x, y)
assert round(abs(cindex - 0.7760582309811175), 7) == 0
+
+ @staticmethod
+ def test_fit_uncomparable(whas500_uncomparable):
+ ssvm = NaiveSurvivalSVM(loss='squared_hinge', dual=False, tol=1e-8, max_iter=1000, random_state=0)
+ with pytest.raises(NoComparablePairException):
+ ssvm.fit(whas500_uncomparable.x, whas500_uncomparable.y)