def testBijector(self):
# TODO(b/152525415): varied/extreme values, forward(+inf/-inf) and inv(0/1)
loc = np.array(0.3, dtype=np.float64)
scale = np.array(5., dtype=np.float64)
concentration = np.array(2., dtype=np.float64)
bijector = tfb.FrechetCDF(loc=loc, scale=scale, concentration=concentration,
validate_args=True)
self.assertStartsWith(bijector.name, 'frechet')
# Frechet distribution
frechet_dist = stats.invweibull(c=concentration, loc=loc, scale=scale)
x = np.array([[[0.3001], [0.8], [2.], [4.2], [12.]]], dtype=np.float64)
y = frechet_dist.cdf(x).astype(np.float64)
self.assertAllClose(y, self.evaluate(bijector.forward(x)))
x = np.array([[[0.49], [0.8], [2.], [4.2], [12.]]], dtype=np.float64)
y = frechet_dist.cdf(x).astype(np.float64)
# the below tests fail if x < 0.49
self.assertAllClose(x, self.evaluate(bijector.inverse(y)))
self.assertAllClose(
np.squeeze(frechet_dist.logpdf(x), axis=-1),
self.evaluate(bijector.forward_log_det_jacobian(x, event_ndims=1)))
self.assertAllClose(
self.evaluate(-bijector.inverse_log_det_jacobian(y, event_ndims=1)),
self.evaluate(bijector.forward_log_det_jacobian(x, event_ndims=1)),
rtol=1e-4,
atol=0.)
with self.assertRaisesOpError(r'Forward transformation input.*than `loc`.'):
self.evaluate(bijector.forward(0.1))
with self.assertRaisesOpError(r'Inverse transformation input.* 0.'):
self.evaluate(bijector.inverse(-0.1))
with self.assertRaisesOpError(r'Inverse transformation input.* 1.'):
self.evaluate(bijector.inverse(1.1))
def testBijectiveAndFinite(self):
loc = np.array(-1., np.float64)
bijector = tfb.FrechetCDF(loc=loc, scale=3.0, concentration=2.,
validate_args=True)
x = np.linspace(loc+0.25, 10., num=10).astype(np.float64)
y = np.linspace(0.01, 0.99, num=10).astype(np.float64)
bijector_test_util.assert_bijective_and_finite(
bijector, x, y, eval_func=self.evaluate, event_ndims=0, rtol=1e-3)
def testScalarCongruency(self):
loc = np.array(-1., np.float64)
bijector_test_util.assert_scalar_congruency(tfb.FrechetCDF(
loc=loc, scale=20., concentration=0.5),
lower_x=1.,
upper_x=100.,
eval_func=self.evaluate,
rtol=0.05)
def testVariablesScaleAndconcentration(self):
x = tf.Variable(1.)
y = tf.Variable(1.)
b = tfb.FrechetCDF(loc=0., scale=x, concentration=y, validate_args=True)
self.evaluate(x.initializer)
self.evaluate(y.initializer)
self.assertIs(x, b.scale)
self.assertIs(y, b.concentration)
self.assertEqual((), self.evaluate(b.forward(3.)).shape)
with self.assertRaisesOpError('Argument `scale` must be positive.'):
with tf.control_dependencies([x.assign(-1.)]):
self.evaluate(b.forward(3.))
with self.assertRaisesOpError('Argument `concentration` must be positive.'):
with tf.control_dependencies([x.assign(1), y.assign(-1.)]):
self.evaluate(b.forward(3.))