def score(self, features=None, times=None, censoring=None):
"""Returns the negative log-likelihood of the model, using the current
fitted coefficients on the passed data.
If no data is passed, the negative log-likelihood is computed using the
data used for training.
Parameters
----------
features : `None` or `numpy.ndarray`, shape=(n_samples, n_features)
The features matrix
times : `None` or `numpy.array`, shape = (n_samples,)
Observed times
censoring : `None` or `numpy.array`, shape = (n_samples,)
Indicator of censoring of each sample.
``True`` means true failure, namely non-censored time.
dtype must be unsigned short
Returns
-------
output : `float`
The value of the negative log-likelihood
"""
if self._fitted:
all_none = all(e is None for e in [features, times, censoring])
if all_none:
return self._model_obj.loss(self.coeffs)
else:
if features is None:
raise ValueError('Passed ``features`` is None')
elif times is None:
raise ValueError('Passed ``times`` is None')
elif censoring is None:
raise ValueError('Passed ``censoring`` is None')
else:
features, times, censoring = self._all_safe(
features, times, censoring)
model = ModelCoxRegPartialLik().fit(
features, times, censoring)
return model.loss(self.coeffs)
else:
raise RuntimeError('You must fit the model first')
def test_set_model(self):
"""...Test set_model of saga, should only accept childs of
ModelGeneralizedLinear"""
# We try to pass a ModelCoxRegPartialLik which is not a generalized
# linear model to SAGA to check that the error is raised
msg = '^SAGA accepts only childs of `ModelGeneralizedLinear`$'
with self.assertRaisesRegex(ValueError, msg):
w = weights_sparse_gauss(n_weights=2, nnz=0)
X, T, C = SimuCoxReg(w).simulate()
model = ModelCoxRegPartialLik().fit(X, T, C)
SAGA().set_model(model)
msg = '^SAGA accepts only childs of `ModelGeneralizedLinear`$'
with self.assertRaisesRegex(RuntimeError, msg):
w = weights_sparse_gauss(n_weights=2, nnz=0)
X, T, C = SimuCoxReg(w).simulate()
model = ModelCoxRegPartialLik().fit(X, T, C)
saga = SAGA()
saga._solver.set_model(model._model)
def test_ModelCoxRegPartialLik(self):
"""...Numerical consistency check of loss and gradient for Cox Regression
"""
np.random.seed(123)
n_samples, n_features = 100, 5
w0 = np.random.randn(n_features)
features, times, censoring = SimuCoxReg(w0, n_samples=n_samples,
verbose=False,
seed=1234).simulate()
model = ModelCoxRegPartialLik()
model.fit(features, times, censoring)
model_spars = ModelCoxRegPartialLik()
model_spars.fit(csr_matrix(features), times, censoring)
self.run_test_for_glm(model, model_spars, 1e-5, 1e-4)
def _construct_model_obj(self):
return ModelCoxRegPartialLik()