Skip to content

Commit

Permalink
Merge branch 'pr/711' into dev
Browse files Browse the repository at this point in the history
  • Loading branch information
@ankurankan
ankurankan committed Jul 22, 2016
2 parents 7ce339b + 09800c9 commit 57fa938
Show file tree
Hide file tree
Showing 6 changed files with 102 additions and 183 deletions.
41 changes: 20 additions & 21 deletions pgmpy/inference/continuous/base.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@ class BaseGradLogPDF(object):
Examples
--------
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.factors import JointGaussianDistribution
>>> from pgmpy.inference.continuous import BaseGradLogPDF
>>> import numpy as np
>>> class GradLogGaussian(BaseGradLogPDF):
Expand Down Expand Up @@ -73,7 +73,7 @@ def get_gradient_log_pdf(self):
--------
>>> # Using implementation of GradLogPDFGaussian
>>> from pgmpy.inference.continuous import GradLogPDFGaussian
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.factors import JointGaussianDistribution
>>> import numpy as np
>>> mean = np.array([1, 1])
>>> covariance = np.array([[1, -5], [-5, 2]])
Expand Down Expand Up @@ -104,7 +104,7 @@ class GradLogPDFGaussian(BaseGradLogPDF):
Example
-------
>>> from pgmpy.inference.continuous import GradLogPDFGaussian
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.factors import JointGaussianDistribution
>>> import numpy as np
>>> mean = np.array([3, 4])
>>> covariance = np.array([[5, 4], [4, 5]])
Expand Down Expand Up @@ -153,9 +153,8 @@ class BaseSimulateHamiltonianDynamics(object):
stepsize: Float
stepsize for the simulating dynamics
grad_log_pdf : A subclass of pgmpy.inference.continuous.BaseGradLogPDF, defaults to None
grad_log_pdf : A subclass of pgmpy.inference.continuous.BaseGradLogPDF
A class for finding gradient log and log of distribution
If none, then model.get_gradient_log_pdf will be used
grad_log_position: A 1d array like object, defaults to None
Vector representing gradient log at given position
Expand All @@ -164,7 +163,7 @@ class BaseSimulateHamiltonianDynamics(object):
Examples
--------
>>> from pgmpy.inference.continuous import BaseSimulateHamiltonianDynamics
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.factors import JointGaussianDistribution
>>> from pgmpy.inference.continuous import GradLogPDFGaussian
>>> import numpy as np
>>> # Class should initalize self.new_position, self.new_momentum and self.new_grad_logp
Expand All @@ -177,7 +176,7 @@ class BaseSimulateHamiltonianDynamics(object):
... def _get_proposed_values(self):
... momentum_bar = self.momentum + self.stepsize * self.grad_log_position
... position_bar = self.position + self.stepsize * momentum_bar
... grad_log_position, _ = self.model.get_gradient_log_pdf(position_bar, self.grad_log_pdf)
... grad_log_position, _ = self.grad_log_pdf(position_bar, self.model).get_gradient_log_pdf()
... return position_bar, momentum_bar, grad_log_position
>>> pos = np.array([1, 2])
>>> momentum = np.array([0, 0])
Expand All @@ -194,21 +193,21 @@ class BaseSimulateHamiltonianDynamics(object):
array([-0.9375, -1.875])
"""

def __init__(self, model, position, momentum, stepsize, grad_log_pdf=None, grad_log_position=None):
def __init__(self, model, position, momentum, stepsize, grad_log_pdf, grad_log_position=None):

position = _check_1d_array_object(position, 'position')

momentum = _check_1d_array_object(momentum, 'momentum')

if grad_log_pdf is not None and not issubclass(grad_log_pdf, BaseGradLogPDF):
if not issubclass(grad_log_pdf, BaseGradLogPDF):
raise TypeError("grad_log_pdf must be an instance" +
" of pgmpy.inference.continuous.base.BaseGradLogPDF")

_check_length_equal(position, momentum, 'position', 'momentum')
_check_length_equal(position, model.variables, 'position', 'model.variables')

if grad_log_position is None:
grad_log_position, _ = model.get_gradient_log_pdf(position, grad_log_pdf)
grad_log_position, _ = grad_log_pdf(position, model).get_gradient_log_pdf()

else:
grad_log_positon = _check_1d_array_object(grad_log_position, 'grad_log_position')
Expand Down Expand Up @@ -242,15 +241,15 @@ def get_proposed_values(self):
Example
-------
>>> # Using implementation of ModifiedEuler
>>> from pgmpy.inference.continuous import ModifiedEuler
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.inference.continuous import ModifiedEuler, GradLogPDFGaussian as GLPG
>>> from pgmpy.factors import JointGaussianDistribution
>>> import numpy as np
>>> pos = np.array([3, 4])
>>> momentum = np.array([1, 1])
>>> mean = np.array([-1, 1])
>>> covariance = 3*np.eye(2)
>>> model = JointGaussianDistribution(['x', 'y'], mean, covariance)
>>> new_pos, new_momentum, new_grad = ModifiedEuler(model, pos, momentum, 0.70).get_proposed_values()
>>> new_pos, new_momentum, new_grad = ModifiedEuler(model, pos, momentum, 0.70, GLPG).get_proposed_values()
>>> new_pos
array([ 3.04666667, 4.21 ])
>>> new_momentum
Expand Down Expand Up @@ -290,15 +289,15 @@ class LeapFrog(BaseSimulateHamiltonianDynamics):
Example
--------
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.inference.continuous import LeapFrog
>>> from pgmpy.factors import JointGaussianDistribution
>>> from pgmpy.inference.continuous import LeapFrog, GradLogPDFGaussian as GLPG
>>> import numpy as np
>>> pos = np.array([2, 1])
>>> momentum = np.array([7, 7])
>>> mean = np.array([-5, 5])
>>> covariance = np.array([[1, 2], [2, 1]])
>>> model = JointGaussianDistribution(['x', 'y'], mean, covariance)
>>> new_pos, new_momentum, new_grad = LeapFrog(model, pos, momentum, 4.0).get_proposed_values()
>>> new_pos, new_momentum, new_grad = LeapFrog(model, pos, momentum, 4.0, GLPG).get_proposed_values()
>>> new_pos
array([ 70., -19.])
>>> new_momentum
Expand All @@ -307,7 +306,7 @@ class LeapFrog(BaseSimulateHamiltonianDynamics):
array([ 41., -58.])
"""

def __init__(self, model, position, momentum, stepsize, grad_log_pdf=None, grad_log_position=None):
def __init__(self, model, position, momentum, stepsize, grad_log_pdf, grad_log_position=None):

BaseSimulateHamiltonianDynamics.__init__(self, model, position, momentum,
stepsize, grad_log_pdf, grad_log_position)
Expand All @@ -324,7 +323,7 @@ def _get_proposed_values(self):
# Take full step in time for updating position position
position_bar = self.position + self.stepsize * momentum_bar

grad_log, _ = self.model.get_gradient_log_pdf(position_bar, self.grad_log_pdf)
grad_log, _ = self.grad_log_pdf(position_bar, self.model).get_gradient_log_pdf()

# Take remaining half step in time for updating momentum
momentum_bar = momentum_bar + 0.5 * self.stepsize * grad_log
Expand Down Expand Up @@ -361,7 +360,7 @@ class ModifiedEuler(BaseSimulateHamiltonianDynamics):
Example
--------
>>> from pgmpy.models import JointGaussianDistribution
>>> from pgmpy.factors import JointGaussianDistribution
>>> from pgmpy.inference.continuous import GradLogPDFGaussian, ModifiedEuler
>>> import numpy as np
>>> pos = np.array([2, 1])
Expand All @@ -379,7 +378,7 @@ class ModifiedEuler(BaseSimulateHamiltonianDynamics):
array([-2.125, -1.1875])
"""

def __init__(self, model, position, momentum, stepsize, grad_log_pdf=None, grad_log_position=None):
def __init__(self, model, position, momentum, stepsize, grad_log_pdf, grad_log_position=None):

BaseSimulateHamiltonianDynamics.__init__(self, model, position, momentum,
stepsize, grad_log_pdf, grad_log_position)
Expand All @@ -396,6 +395,6 @@ def _get_proposed_values(self):
# Take full step in time and update position
position_bar = self.position + self.stepsize * momentum_bar

grad_log, _ = self.model.get_gradient_log_pdf(position_bar, self.grad_log_pdf)
grad_log, _ = self.grad_log_pdf(position_bar, self.model).get_gradient_log_pdf()

return position_bar, momentum_bar, grad_log
Loading

0 comments on commit 57fa938

Please sign in to comment.