def test__trasnform(self):
# setup
instance = MagicMock()
trials = np.array([[0.1, 0.8, 0.2], [0.7, 0.6, 0.4], [0.4, 0.2, 0.3],
[0.7, 0.8, 0.9], [0.2, 0.3, 0.4]])
distributions = []
for column in trials.T:
distribution = Univariate()
distribution.fit(column)
distributions.append(distribution)
instance._distributions = distributions
# run
result = GaussianCopulaProcessMetaModel._transform(instance, trials)
# assert
expected_result = np.array([[-0.93051407, 0.72584367, -1.05823838],
[0.78865964, 0.1155097, 0.3732126],
[-0.04619363, -0.96388991, -0.49278658],
[0.78865964, 0.72584367, 1.82607428],
[-0.60472608, -0.66069297, 0.3732126]])
np.testing.assert_allclose(result, expected_result)
def _fit(self, trials, scores):
self._distributions = []
for column in trials.T:
distribution = Univariate()
distribution.fit(column)
self._distributions.append(distribution)
distribution = Univariate()
distribution.fit(scores)
self._score_distribution = distribution
trans_trials = self._transform(trials)
trans_scores = scipy.stats.norm.ppf(
self._score_distribution.cdf(scores).clip(0 + EPSILON, 1 - EPSILON)
)
super()._fit(trans_trials, trans_scores)