def test_hpe_likelihood_log_likelihood_equal_ratio_zero_evidence(self):
like = HyperparameterLikelihood(
posteriors=self.data, hyper_prior=self.model,
selection_function=self.selection_function,
ln_evidences=self.ln_evidences)
like.parameters.update(self.params)
self.assertEqual(like.log_likelihood_ratio(), like.log_likelihood())
def test_hpe_likelihood_set_selection(self):
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
selection_function=self.selection_function,
)
self.assertEqual(like.selection_function("foo"), 2.0)
def test_hpe_likelihood_set_conversion(self):
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
conversion_function=self.conversion_function,
)
self.assertEqual(like.conversion_function("foo"), ("foo", ["bar"]))
def test_generate_rate_posterior_sample_returns_positive(self):
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
selection_function=self.selection_function,
ln_evidences=self.ln_evidences,
)
self.assertGreater(like.generate_rate_posterior_sample(), 0)
def test_generate_rate_posterior_sample_raises_error(self):
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
selection_function=self.selection_function,
ln_evidences=self.ln_evidences,
)
with self.assertRaises(NotImplementedError):
like.generate_rate_posterior_sample()
def test_hpe_likelihood_conversion_function_pops_parameters(self):
like = HyperparameterLikelihood(
posteriors=self.data, hyper_prior=self.model,
conversion_function=self.conversion_function,
selection_function=self.selection_function,
ln_evidences=self.ln_evidences)
self.params['bar'] = None
like.parameters.update(self.params)
like.log_likelihood_ratio()
self.assertFalse('bar' in like.parameters)
def test_resampling_posteriors(self):
priors = PriorDict(
dict(a=Uniform(0, 2), b=Uniform(0, 2), c=Uniform(0, 2)))
samples = priors.sample(100)
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
selection_function=self.selection_function,
ln_evidences=self.ln_evidences,
)
new_samples = like.posterior_predictive_resample(samples=samples)
for key in new_samples:
self.assertEqual(new_samples[key].shape, like.data[key].shape)
def test_likelihood_pass_sampling_prior_raises_error(self):
with self.assertRaises(TypeError):
_ = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
sampling_prior=lambda dataset: 5,
)
def test_prior_in_posteriors(self):
for frame in self.data:
frame["prior"] = 1
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model)
self.assertTrue(
xp.array_equal(like.sampling_prior, xp.ones_like(like.data["a"])))
def test_generate_extra_statistics(self):
like = HyperparameterLikelihood(
posteriors=self.data,
hyper_prior=self.model,
selection_function=self.selection_function,
conversion_function=self.conversion_function,
ln_evidences=self.ln_evidences,
)
self.params["bar"] = None
new_params = like.generate_extra_statistics(sample=self.params.copy())
expected = {
"a": 1,
"b": 1,
"c": 1,
"ln_bf_0": 0.0,
"ln_bf_1": 0.0,
"ln_bf_2": 0.0,
"ln_bf_3": 0.0,
"ln_bf_4": 0.0,
"selection": 2.0,
"bar": None,
}
self.assertDictEqual(expected, new_params)
def test_hpe_likelihood_dont_set_evidences(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model)
self.assertTrue(xp.isnan(like.total_noise_evidence))
def test_hpe_likelihood_set_evidences(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model,
ln_evidences=self.ln_evidences)
self.assertEqual(like.total_noise_evidence, 0)
def test_not_passing_prior(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model)
self.assertEqual(like.sampling_prior, 1)
def test_likelihood_pass_sampling_prior_works(self):
like = HyperparameterLikelihood(
posteriors=self.data, hyper_prior=self.model,
sampling_prior=lambda x: 5)
self.assertEqual(like.sampling_prior, 5)
def test_hpe_likelihood_log_likelihood_ratio(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model)
like.parameters.update(self.params)
self.assertEqual(like.log_likelihood_ratio(), 0.0)
def test_hpe_likelihood_set_conversion(self):
like = HyperparameterLikelihood(
posteriors=self.data, hyper_prior=self.model,
conversion_function=self.conversion_function)
self.assertEqual(like.conversion_function('foo'), ('foo', ['bar']))
def test_hpe_likelihood_converts_nan_to_neginf(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model)
like._get_selection_factor = lambda *args, **kwargs: np.nan
like.parameters.update(self.params)
self.assertEqual(like.log_likelihood_ratio(), np.nan_to_num(-np.inf))
def test_hpe_likelihood_set_max_samples(self):
like = HyperparameterLikelihood(posteriors=self.data,
hyper_prior=self.model,
max_samples=10)
self.assertEqual(like.data["a"].shape, (5, 10))
def test_hpe_likelihood_requires_posteriors(self):
with self.assertRaises(TypeError):
_ = HyperparameterLikelihood(hyper_prior=self.model)
def get_agn_spin_likelihood(posteriors):
return HyperparameterLikelihood(
posteriors=posteriors,
hyper_prior=agn_spin
)
def test_hpe_likelihood_requires_hyper_prior(self):
with self.assertRaises(TypeError):
_ = HyperparameterLikelihood(posteriors=self.data)
