def test_cauchy_prior(self):
# Test two specific function values
p1 = pints.CauchyLogPrior(0, 10)
self.assertEqual(p1([0]), -3.447314978843446)
p2 = pints.CauchyLogPrior(10, 5)
self.assertTrue(np.abs(p2([10]) + 2.7541677982835) < 0.001)
p3 = pints.CauchyLogPrior(-3, 3.5)
self.assertAlmostEqual(p3([1.4]), -3.3454404435815586)
# Test exceptions
self.assertRaises(ValueError, pints.CauchyLogPrior, 0, 0)
self.assertRaises(ValueError, pints.CauchyLogPrior, 10, -1)
# Test means
self.assertTrue(np.isnan(p1.mean()))
self.assertTrue(np.isnan(p2.mean()))
# Test other function calls
self.assertEqual(p1.n_parameters(), 1)
self.assertEqual(p2.n_parameters(), 1)
# Test sensitivities
p = pints.CauchyLogPrior(10, 5)
val = p([-3.3])
val1, dp = p.evaluateS1([-3.3])
self.assertEqual(val, val1)
self.assertAlmostEqual(dp[0], 0.13175491604338996, places=6)
def test_cauchy_prior_sampling(self):
p1 = pints.CauchyLogPrior(0, 1000)
self.assertEqual(len(p1.sample()), 1)
self.assertEqual(len(p1.sample(100)), 100)
p1 = pints.CauchyLogPrior(-4.7, 10)
n = 1000000
v_samples = p1.sample(n)
self.assertTrue(np.abs(np.median(v_samples) - -4.695) < 0.1)
def test_cauchy_prior(self):
# Test two specific function values
p1 = pints.CauchyLogPrior(0, 10)
self.assertEqual(p1([0]), -3.447314978843446)
p2 = pints.CauchyLogPrior(10, 5)
self.assertTrue(np.abs(p2([10]) + 2.7541677982835) < 0.001)
# Test exceptions
self.assertRaises(ValueError, pints.CauchyLogPrior, 0, 0)
self.assertRaises(ValueError, pints.CauchyLogPrior, 10, -1)
# Test means
self.assertTrue(np.isnan(p1.mean()))
self.assertTrue(np.isnan(p2.mean()))
# Test other function calls
self.assertEqual(p1.n_parameters(), 1)
self.assertEqual(p2.n_parameters(), 1)
def __init__(self, location, scale):
# Test inputs
if float(scale) <= 0:
raise ValueError('Scale must be positive')
self._location = location
self._scale = scale
self._cauchy = pints.CauchyLogPrior(location=self._location,
scale=self._scale)
# Cache constants
self._norm_factor = -np.log(np.arctan(location / scale) / np.pi + 0.5)
def test_cauchy_cdf_icdf(self):
p = pints.CauchyLogPrior(-3, 2)
self.assertAlmostEqual(p.cdf(5.5), 0.92644155602673783)
self.assertAlmostEqual(p.icdf(0.1), -9.1553670743505062)
def test_cauchy_prior_sampling(self):
# Aren't many tests for Cauchy distributions
# because they have no mean or variance!
p1 = pints.CauchyLogPrior(0, 1000)
self.assertEqual(len(p1.sample()), 1)
self.assertEqual(len(p1.sample(100)), 100)
