def _sample_numerical(
self,
low, # type: float
high, # type: float
below, # type: np.ndarray
above, # type: np.ndarray
q=None, # type: Optional[float]
is_log=False # type: bool
):
# type: (...) -> float
if is_log:
low = np.log(low)
high = np.log(high)
below = np.log(below)
above = np.log(above)
size = (self.n_ei_candidates, )
parzen_estimator_below = ParzenEstimator(
mus=below,
low=low,
high=high,
parameters=self.parzen_estimator_parameters)
samples_below = self._sample_from_gmm(
parzen_estimator=parzen_estimator_below,
low=low,
high=high,
q=q,
is_log=is_log,
size=size)
log_likelihoods_below = self._gmm_log_pdf(
samples=samples_below,
parzen_estimator=parzen_estimator_below,
low=low,
high=high,
q=q,
is_log=is_log)
parzen_estimator_above = ParzenEstimator(
mus=above,
low=low,
high=high,
parameters=self.parzen_estimator_parameters)
log_likelihoods_above = self._gmm_log_pdf(
samples=samples_below,
parzen_estimator=parzen_estimator_above,
low=low,
high=high,
q=q,
is_log=is_log)
return float(
TPESampler._compare(samples=samples_below,
log_l=log_likelihoods_below,
log_g=log_likelihoods_above)[0])
def test_calculate(mus, flags, expected):
# type: (List[float], Dict[str, bool], Dict[str, List[float]]) -> None
s_weights, s_mus, s_sigmas = \
ParzenEstimator._calculate(mus, -1.0, 1.0, prior_weight=1.0,
consider_prior=flags['prior'],
consider_magic_clip=flags['magic_clip'],
consider_endpoints=flags['endpoints'],
weights_func=default_weights)
# Result contains an additional value for a prior distribution if consider_prior is True.
np.testing.assert_almost_equal(s_weights, expected['weights'])
np.testing.assert_almost_equal(s_mus, expected['mus'])
np.testing.assert_almost_equal(s_sigmas, expected['sigmas'])
def test_calculate_shape_check(mus, prior, magic_clip, endpoints):
# type: (List[float], bool, bool, bool) -> None
s_weights, s_mus, s_sigmas = \
ParzenEstimator._calculate(mus, -1.0, 1.0, prior_weight=1.0,
consider_prior=prior,
consider_magic_clip=magic_clip,
consider_endpoints=endpoints,
weights_func=default_weights)
# Result contains an additional value for a prior distribution if prior is True.
assert len(s_weights) == len(mus) + int(prior)
assert len(s_mus) == len(mus) + int(prior)
assert len(s_sigmas) == len(mus) + int(prior)
