コード例 #1
0
 def setUp(self):
     mean = np.array([-1, 1, -1])
     covariance = np.array([[1, 0.6, 0.5], [0.6, 2, 0.3], [0.5, 0.3, 1]])
     self.test_model = JGD(['x', 'y', 'z'], mean, covariance)
     position = [0, 0, 0]
     momentum = [-1, -1, -1]
     self.test_with_grad_log = LeapFrog(model=self.test_model, position=position, momentum=momentum,
                                        stepsize=0.3, grad_log_pdf=GradLogPDFGaussian, grad_log_position=None)
     grad_log_position, _ = GradLogPDFGaussian(position, self.test_model).get_gradient_log_pdf()
     self.test_without_grad_log = LeapFrog(model=self.test_model, position=position, momentum=momentum,
                                           stepsize=0.4, grad_log_pdf=GradLogPDFGaussian,
                                           grad_log_position=grad_log_position)
コード例 #2
0
class TestLeapFrog(unittest.TestCase):

    def setUp(self):
        mean = np.array([-1, 1, -1])
        covariance = np.array([[1, 0.6, 0.5], [0.6, 2, 0.3], [0.5, 0.3, 1]])
        self.test_model = JGD(['x', 'y', 'z'], mean, covariance)
        position = [0, 0, 0]
        momentum = [-1, -1, -1]
        self.test_with_grad_log = LeapFrog(model=self.test_model, position=position, momentum=momentum,
                                           stepsize=0.3, grad_log_pdf=GradLogPDFGaussian, grad_log_position=None)
        grad_log_position, _ = GradLogPDFGaussian(position, self.test_model).get_gradient_log_pdf()
        self.test_without_grad_log = LeapFrog(model=self.test_model, position=position, momentum=momentum,
                                              stepsize=0.4, grad_log_pdf=GradLogPDFGaussian,
                                              grad_log_position=grad_log_position)

    def test_errors(self):
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model, position=1, momentum=[1, 1], stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model, position=[1, 1], momentum=1, stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model, position=[1, 1], momentum=[1], stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model, position=[1], momentum=[1], stepsize=0.1, grad_log_pdf=1)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model, position=[1, 1], momentum=[1, 1], stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model, position=[1, 1, 1], momentum=[1, 1, 1], stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian, grad_log_position=1)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model, position=[1, 1, 1], momentum=[1, 1, 1], stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian, grad_log_position=[1, 1])

    def test_leapfrog_methods(self):
        new_pos, new_momentum, new_grad = self.test_with_grad_log.get_proposed_values()
        np.testing.assert_almost_equal(new_pos, np.array([-0.35634146, -0.25609756, -0.33]))
        np.testing.assert_almost_equal(new_momentum, np.array([-1.3396624, -0.70344884, -1.16963415]))
        np.testing.assert_almost_equal(new_grad, np.array([-1.0123835, 1.00139798, -0.46422764]))
        new_pos, new_momentum, new_grad = self.test_without_grad_log.get_proposed_values()
        np.testing.assert_almost_equal(new_pos, np.array([-0.5001626, -0.32195122, -0.45333333]))
        np.testing.assert_almost_equal(new_momentum, np.array([-1.42947981, -0.60709102, -1.21246612]))
        np.testing.assert_almost_equal(new_grad, np.array([-0.89536651, 0.98893516, -0.39566396]))

    def tearDown(self):
        del self.test_model
        del self.test_with_grad_log
        del self.test_without_grad_log
コード例 #3
0
 def test_errors(self):
     with self.assertRaises(TypeError):
         LeapFrog(
             model=self.test_model,
             position=1,
             momentum=[1, 1],
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
         )
     with self.assertRaises(TypeError):
         LeapFrog(
             model=self.test_model,
             position=[1, 1],
             momentum=1,
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
         )
     with self.assertRaises(ValueError):
         LeapFrog(
             model=self.test_model,
             position=[1, 1],
             momentum=[1],
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
         )
     with self.assertRaises(TypeError):
         LeapFrog(
             model=self.test_model,
             position=[1],
             momentum=[1],
             stepsize=0.1,
             grad_log_pdf=1,
         )
     with self.assertRaises(ValueError):
         LeapFrog(
             model=self.test_model,
             position=[1, 1],
             momentum=[1, 1],
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
         )
     with self.assertRaises(TypeError):
         LeapFrog(
             model=self.test_model,
             position=[1, 1, 1],
             momentum=[1, 1, 1],
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
             grad_log_position=1,
         )
     with self.assertRaises(ValueError):
         LeapFrog(
             model=self.test_model,
             position=[1, 1, 1],
             momentum=[1, 1, 1],
             stepsize=0.1,
             grad_log_pdf=GradLogPDFGaussian,
             grad_log_position=[1, 1],
         )
コード例 #4
0
 def setUp(self):
     mean = np.array([-1, 1, -1])
     covariance = np.array([[1, 0.6, 0.5], [0.6, 2, 0.3], [0.5, 0.3, 1]])
     self.test_model = JGD(['x', 'y', 'z'], mean, covariance)
     position = [0, 0, 0]
     momentum = [-1, -1, -1]
     self.test_with_grad_log = LeapFrog(model=self.test_model,
                                        position=position,
                                        momentum=momentum,
                                        stepsize=0.3,
                                        grad_log_pdf=GradLogPDFGaussian,
                                        grad_log_position=None)
     grad_log_position, _ = GradLogPDFGaussian(
         position, self.test_model).get_gradient_log_pdf()
     self.test_without_grad_log = LeapFrog(
         model=self.test_model,
         position=position,
         momentum=momentum,
         stepsize=0.4,
         grad_log_pdf=GradLogPDFGaussian,
         grad_log_position=grad_log_position)
コード例 #5
0
class TestLeapFrog(unittest.TestCase):
    def setUp(self):
        mean = np.array([-1, 1, -1])
        covariance = np.array([[1, 0.6, 0.5], [0.6, 2, 0.3], [0.5, 0.3, 1]])
        self.test_model = JGD(['x', 'y', 'z'], mean, covariance)
        position = [0, 0, 0]
        momentum = [-1, -1, -1]
        self.test_with_grad_log = LeapFrog(model=self.test_model,
                                           position=position,
                                           momentum=momentum,
                                           stepsize=0.3,
                                           grad_log_pdf=GradLogPDFGaussian,
                                           grad_log_position=None)
        grad_log_position, _ = GradLogPDFGaussian(
            position, self.test_model).get_gradient_log_pdf()
        self.test_without_grad_log = LeapFrog(
            model=self.test_model,
            position=position,
            momentum=momentum,
            stepsize=0.4,
            grad_log_pdf=GradLogPDFGaussian,
            grad_log_position=grad_log_position)

    def test_errors(self):
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model,
                     position=1,
                     momentum=[1, 1],
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model,
                     position=[1, 1],
                     momentum=1,
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model,
                     position=[1, 1],
                     momentum=[1],
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model,
                     position=[1],
                     momentum=[1],
                     stepsize=0.1,
                     grad_log_pdf=1)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model,
                     position=[1, 1],
                     momentum=[1, 1],
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian)
        with self.assertRaises(TypeError):
            LeapFrog(model=self.test_model,
                     position=[1, 1, 1],
                     momentum=[1, 1, 1],
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian,
                     grad_log_position=1)
        with self.assertRaises(ValueError):
            LeapFrog(model=self.test_model,
                     position=[1, 1, 1],
                     momentum=[1, 1, 1],
                     stepsize=0.1,
                     grad_log_pdf=GradLogPDFGaussian,
                     grad_log_position=[1, 1])

    def test_leapfrog_methods(self):
        new_pos, new_momentum, new_grad = self.test_with_grad_log.get_proposed_values(
        )
        np.testing.assert_almost_equal(
            new_pos, np.array([-0.35634146, -0.25609756, -0.33]))
        np.testing.assert_almost_equal(
            new_momentum, np.array([-1.3396624, -0.70344884, -1.16963415]))
        np.testing.assert_almost_equal(
            new_grad, np.array([-1.0123835, 1.00139798, -0.46422764]))
        new_pos, new_momentum, new_grad = self.test_without_grad_log.get_proposed_values(
        )
        np.testing.assert_almost_equal(
            new_pos, np.array([-0.5001626, -0.32195122, -0.45333333]))
        np.testing.assert_almost_equal(
            new_momentum, np.array([-1.42947981, -0.60709102, -1.21246612]))
        np.testing.assert_almost_equal(
            new_grad, np.array([-0.89536651, 0.98893516, -0.39566396]))

    def tearDown(self):
        del self.test_model
        del self.test_with_grad_log
        del self.test_without_grad_log