def test_iid_normal_passes(self):
n_samples = 500
# five scalar chains taken from iid Normal(0, 1)
rng = test_util.test_np_rng()
iid_normal_samples = rng.randn(n_samples, 5)
rhat_reducer = tfp.experimental.mcmc.PotentialScaleReductionReducer(
independent_chain_ndims=1)
rhat = self.evaluate(test_fixtures.reduce(rhat_reducer, iid_normal_samples))
self.assertAllEqual((), rhat.shape)
self.assertAllClose(1., rhat, rtol=0.02)
def test_offset_normal_fails(self):
n_samples = 500
# three 4-variate chains taken from Normal(0, 1) that have been
# shifted. Since every chain is shifted, they are not the same, and the
# test should fail.
offset = np.array([1., -1., 2.]).reshape(3, 1)
rng = test_util.test_np_rng()
offset_samples = rng.randn(n_samples, 3, 4) + offset
rhat_reducer = tfp.experimental.mcmc.PotentialScaleReductionReducer(
independent_chain_ndims=1)
rhat = self.evaluate(test_fixtures.reduce(rhat_reducer, offset_samples))
self.assertAllEqual((4,), rhat.shape)
self.assertAllGreater(rhat, 1.2)
def test_random_sanity_check(self): rng = test_util.test_np_rng() x = rng.rand(100) cov_reducer = tfp.experimental.mcmc.CovarianceReducer() final_cov = self.evaluate(test_fixtures.reduce(cov_reducer, x)) self.assertNear(np.var(x, ddof=0), final_cov, err=1e-6)
