| | |
- builtins.object
-
- Distribution
-
- MCMCDistribution
-
- OPAD
- TrueDistribution
class Distribution(builtins.object) |
| |
Distribution(particles: Iterable = [], logp: Iterable = [], theta: Dict = None)
Base class for distributions.
|
| |
Methods defined here:
- __add__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Add two distributions together by combining their particles and frequencies.
Parameters:
other (Distribution): The distribution to add.
Returns:
Distribution: The resulting distribution after addition.
- __contains__(self, particle)
- Checks if particle is in the distribution.
- __copy__(self)
- Create a copy of the current distribution.
Returns:
Distribution: A copy of the distribution.
- __init__(self, particles: Iterable = [], logp: Iterable = [], theta: Dict = None)
- Initialise distribution. Particles are stored internally as a dictionary to store their information.
Parameters:
particles (Iterable): List of particles to add in the distribution
logp (Iterable): Scores (log probabilities) of the particles
theta (Dict): Additional particles data
- __len__(self)
- Get the number of particles in the distribution.
Returns:
int: The number of particles.
- __sub__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Subtract one distribution from another by removing particles present in the other distribution.
Parameters:
other (Distribution): The distribution to subtract.
Returns:
Distribution: The resulting distribution after subtraction.
- clear(self)
- Clear all particles from the distribution.
- copy(self)
- hist(self, prop='freq', normalise=False)
- Returns the histogram of particles.
Parameters:
normalise (bool): If True, return normalised counts.
- normalise(self, prop='freq', log=False)
- Normalise the current set of particles in the distribution.
- plot(self, prop='freq', sort=True, normalise=False, limit=-1, ax=None, showxticklabels=False)
- Plot a histogram of the particles in the distribution.
Parameters:
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
tuple: Bars, particles, and counts displayed in the plot.
- prop(self, name)
- Return data about particles.
Parameter:
name: Key name for the particle data
Returns:
(list) particle data
- save(self, filename: str, compression='gzip')
- Saves the Graph object to a file.
Parameters:
filename (str): Path to the output file.
- top(self, prop='freq', n=1)
- Retrieve the top N particles based on a specified property.
Parameters:
prop (str): The property to sort by (default is 'freq').
n (int): The number of top particles to retrieve.
Returns:
np.ndarray: Array of the top N particles.
- update(self, particle, data, **kwargs)
- Adds particle to distribution. If particle already exists, update its data.
Parameter:
particle (Hashable): Particle to add
data (dict): Particle data
Class methods defined here:
- compute_distribution(data: pandas.core.frame.DataFrame, score: structure_learning.scores.score.Score, graph_type='dag', blocklist: numpy.ndarray = None) -> Type[ForwardRef('D')]
- Compute a distribution from data and a scoring function.
Parameters:
data (pd.DataFrame): The dataset to compute the distribution from.
score (Score): The scoring function to evaluate particles.
graph_type (str): The type of graph to use ('dag' or 'cpdag').
Returns:
Distribution: The computed distribution.
- load(filename: str, compression='gzip')
- Loads a Graph object from a file.
Parameters:
filename (str): Path to the input file.
Returns:
Graph: Loaded Graph object.
- plot_multiple(dists: List[Type[ForwardRef('D')]], prop='freq', normalise=False, limit=-1, ax=None, labels=None)
- Plot multiple distributions on the same axes.
Parameters:
dists (List[Distribution]): List of distributions to plot.
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
list: List of bar containers for each distribution.
Readonly properties defined here:
- logp
- Retrieve the log probabilities of all particles in the distribution.
Returns:
np.ndarray: Array of log probabilities.
Data descriptors defined here:
- __dict__
- dictionary for instance variables
- __weakref__
- list of weak references to the object
|
class MCMCDistribution(Distribution) |
| |
MCMCDistribution(particles: Iterable = [], logp: Iterable = [], theta: Dict = None, keep_rejected: bool = True)
|
| |
- Method resolution order:
- MCMCDistribution
- Distribution
- builtins.object
Methods defined here:
- __copy__(self)
- Create a shallow copy of the current distribution.
Returns:
Distribution: A shallow copy of the distribution.
- __init__(self, particles: Iterable = [], logp: Iterable = [], theta: Dict = None, keep_rejected: bool = True)
- Initialise MCMC distribution.
Parameters:
particles (Iterable): List of particles to add in the distribution
logp (Iterable): Scores (log probabilities) of the particles
theta (Dict): Additional particles data
keep_rejected (bool): Store rejected particles
- to_iterates(self)
- Convert an MCMCDistribution to iteration data.
Returns:
dict: A dictionary where keys are iteration numbers and values are data about the particles.
- to_opad(self, plus=False)
- Convert the current MCMC distribution to an OPAD distribution.
Parameters:
plus (bool): If True, include rejected particles in the OPAD distribution.
Returns:
OPAD: The OPAD distribution.
- update(self, particle, data, iteration, **kwargs)
- Adds particle to distribution. If particle already exists, update its data.
Parameter:
particle (Hashable): Particle to add
iteration (int): Iteration number at which the particle was generated.
data (dict): Particle data
Class methods defined here:
- from_iterates(iterates: dict)
- Create an MCMCDistribution from iteration data.
Parameters:
iterates (dict): A dictionary where keys are iteration numbers and values are data about the particles.
Returns:
MCMCDistribution: The resulting MCMC distribution.
Methods inherited from Distribution:
- __add__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Add two distributions together by combining their particles and frequencies.
Parameters:
other (Distribution): The distribution to add.
Returns:
Distribution: The resulting distribution after addition.
- __contains__(self, particle)
- Checks if particle is in the distribution.
- __len__(self)
- Get the number of particles in the distribution.
Returns:
int: The number of particles.
- __sub__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Subtract one distribution from another by removing particles present in the other distribution.
Parameters:
other (Distribution): The distribution to subtract.
Returns:
Distribution: The resulting distribution after subtraction.
- clear(self)
- Clear all particles from the distribution.
- copy(self)
- hist(self, prop='freq', normalise=False)
- Returns the histogram of particles.
Parameters:
normalise (bool): If True, return normalised counts.
- normalise(self, prop='freq', log=False)
- Normalise the current set of particles in the distribution.
- plot(self, prop='freq', sort=True, normalise=False, limit=-1, ax=None, showxticklabels=False)
- Plot a histogram of the particles in the distribution.
Parameters:
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
tuple: Bars, particles, and counts displayed in the plot.
- prop(self, name)
- Return data about particles.
Parameter:
name: Key name for the particle data
Returns:
(list) particle data
- save(self, filename: str, compression='gzip')
- Saves the Graph object to a file.
Parameters:
filename (str): Path to the output file.
- top(self, prop='freq', n=1)
- Retrieve the top N particles based on a specified property.
Parameters:
prop (str): The property to sort by (default is 'freq').
n (int): The number of top particles to retrieve.
Returns:
np.ndarray: Array of the top N particles.
Class methods inherited from Distribution:
- compute_distribution(data: pandas.core.frame.DataFrame, score: structure_learning.scores.score.Score, graph_type='dag', blocklist: numpy.ndarray = None) -> Type[ForwardRef('D')]
- Compute a distribution from data and a scoring function.
Parameters:
data (pd.DataFrame): The dataset to compute the distribution from.
score (Score): The scoring function to evaluate particles.
graph_type (str): The type of graph to use ('dag' or 'cpdag').
Returns:
Distribution: The computed distribution.
- load(filename: str, compression='gzip')
- Loads a Graph object from a file.
Parameters:
filename (str): Path to the input file.
Returns:
Graph: Loaded Graph object.
- plot_multiple(dists: List[Type[ForwardRef('D')]], prop='freq', normalise=False, limit=-1, ax=None, labels=None)
- Plot multiple distributions on the same axes.
Parameters:
dists (List[Distribution]): List of distributions to plot.
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
list: List of bar containers for each distribution.
Readonly properties inherited from Distribution:
- logp
- Retrieve the log probabilities of all particles in the distribution.
Returns:
np.ndarray: Array of log probabilities.
Data descriptors inherited from Distribution:
- __dict__
- dictionary for instance variables
- __weakref__
- list of weak references to the object
|
class OPAD(MCMCDistribution) |
| |
OPAD(particles: Iterable = [], logp: Iterable = [], theta: Dict = [], plus=False)
This class implements the OPAD re-weighing mechanism described in
|
| |
- Method resolution order:
- OPAD
- MCMCDistribution
- Distribution
- builtins.object
Methods defined here:
- __copy__(self)
- Create a copy of the current distribution.
Returns:
Distribution: A copy of the distribution.
- __init__(self, particles: Iterable = [], logp: Iterable = [], theta: Dict = [], plus=False)
- Initialise MCMC distribution.
Parameters:
particles (Iterable): List of particles to add in the distribution
logp (Iterable): Scores (log probabilities) of the particles
theta (Dict): Additional particles data
keep_rejected (bool): Store rejected particles
- normalise(self)
- Normalise the OPAD distribution by adding rejected particles and computing probabilities.
Returns:
OPAD: The normalised OPAD distribution.
- plot(self, prop='p', sort=True, normalise=False, limit=-1)
- Plot a histogram of the particles in the OPAD distribution.
Parameters:
prop (str): The property to plot (default is 'p').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
Returns:
tuple: Bars, particles, and counts displayed in the plot.
- to_iterates(self)
- Convert an MCMCDistribution to iteration data.
Returns:
dict: A dictionary where keys are iteration numbers and values are data about the particles.
- update(self, particle, data, iteration, normalise=True)
- Add new particles to the OPAD distribution and optionally renormalise.
Parameters:
particle (Hashable): The particle to add.
iteration (int): The iteration number at which the particle was generated.
data (dict): Data associated with the particle.
normalise (bool): If True, renormalise the distribution after adding the particle.
Class methods defined here:
- compute_normalisation(logp: Union[List, numpy.ndarray], return_constants=True)
- Compute the normalisation factor given the log scores.
Parameters:
logp (list | np.ndarray): The log scores
return_constants (bool): If True, also returns log(Z) and max score.
Returns:
(np.array): Normalised scores
(float): Normalisation factor
(np.array): Maximum score
- from_mcmc(dist: structure_learning.distributions.distribution.Distribution, plus=False)
- Create an OPAD distribution from an MCMC distribution.
Parameters:
dist (Distribution): The MCMC distribution to convert.
plus (bool): Whether to include rejected particles in the OPAD distribution.
Returns:
OPAD: The resulting OPAD distribution.
Methods inherited from MCMCDistribution:
- to_opad(self, plus=False)
- Convert the current MCMC distribution to an OPAD distribution.
Parameters:
plus (bool): If True, include rejected particles in the OPAD distribution.
Returns:
OPAD: The OPAD distribution.
Class methods inherited from MCMCDistribution:
- from_iterates(iterates: dict)
- Create an MCMCDistribution from iteration data.
Parameters:
iterates (dict): A dictionary where keys are iteration numbers and values are data about the particles.
Returns:
MCMCDistribution: The resulting MCMC distribution.
Methods inherited from Distribution:
- __add__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Add two distributions together by combining their particles and frequencies.
Parameters:
other (Distribution): The distribution to add.
Returns:
Distribution: The resulting distribution after addition.
- __contains__(self, particle)
- Checks if particle is in the distribution.
- __len__(self)
- Get the number of particles in the distribution.
Returns:
int: The number of particles.
- __sub__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Subtract one distribution from another by removing particles present in the other distribution.
Parameters:
other (Distribution): The distribution to subtract.
Returns:
Distribution: The resulting distribution after subtraction.
- clear(self)
- Clear all particles from the distribution.
- copy(self)
- hist(self, prop='freq', normalise=False)
- Returns the histogram of particles.
Parameters:
normalise (bool): If True, return normalised counts.
- prop(self, name)
- Return data about particles.
Parameter:
name: Key name for the particle data
Returns:
(list) particle data
- save(self, filename: str, compression='gzip')
- Saves the Graph object to a file.
Parameters:
filename (str): Path to the output file.
- top(self, prop='freq', n=1)
- Retrieve the top N particles based on a specified property.
Parameters:
prop (str): The property to sort by (default is 'freq').
n (int): The number of top particles to retrieve.
Returns:
np.ndarray: Array of the top N particles.
Class methods inherited from Distribution:
- compute_distribution(data: pandas.core.frame.DataFrame, score: structure_learning.scores.score.Score, graph_type='dag', blocklist: numpy.ndarray = None) -> Type[ForwardRef('D')]
- Compute a distribution from data and a scoring function.
Parameters:
data (pd.DataFrame): The dataset to compute the distribution from.
score (Score): The scoring function to evaluate particles.
graph_type (str): The type of graph to use ('dag' or 'cpdag').
Returns:
Distribution: The computed distribution.
- load(filename: str, compression='gzip')
- Loads a Graph object from a file.
Parameters:
filename (str): Path to the input file.
Returns:
Graph: Loaded Graph object.
- plot_multiple(dists: List[Type[ForwardRef('D')]], prop='freq', normalise=False, limit=-1, ax=None, labels=None)
- Plot multiple distributions on the same axes.
Parameters:
dists (List[Distribution]): List of distributions to plot.
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
list: List of bar containers for each distribution.
Readonly properties inherited from Distribution:
- logp
- Retrieve the log probabilities of all particles in the distribution.
Returns:
np.ndarray: Array of log probabilities.
Data descriptors inherited from Distribution:
- __dict__
- dictionary for instance variables
- __weakref__
- list of weak references to the object
|
class TrueDistribution(Distribution) |
| |
TrueDistribution(particles=[], logp=[], theta=None)
|
| |
- Method resolution order:
- TrueDistribution
- Distribution
- builtins.object
Methods defined here:
- __init__(self, particles=[], logp=[], theta=None)
- Initialise distribution. Particles are stored internally as a dictionary to store their information.
Parameters:
particles (Iterable): List of particles to add in the distribution
logp (Iterable): Scores (log probabilities) of the particles
theta (Dict): Additional particles data
- normalise(self)
- Normalise the current set of particles in the distribution.
Methods inherited from Distribution:
- __add__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Add two distributions together by combining their particles and frequencies.
Parameters:
other (Distribution): The distribution to add.
Returns:
Distribution: The resulting distribution after addition.
- __contains__(self, particle)
- Checks if particle is in the distribution.
- __copy__(self)
- Create a copy of the current distribution.
Returns:
Distribution: A copy of the distribution.
- __len__(self)
- Get the number of particles in the distribution.
Returns:
int: The number of particles.
- __sub__(self, other: Type[ForwardRef('D')]) -> Type[ForwardRef('D')]
- Subtract one distribution from another by removing particles present in the other distribution.
Parameters:
other (Distribution): The distribution to subtract.
Returns:
Distribution: The resulting distribution after subtraction.
- clear(self)
- Clear all particles from the distribution.
- copy(self)
- hist(self, prop='freq', normalise=False)
- Returns the histogram of particles.
Parameters:
normalise (bool): If True, return normalised counts.
- plot(self, prop='freq', sort=True, normalise=False, limit=-1, ax=None, showxticklabels=False)
- Plot a histogram of the particles in the distribution.
Parameters:
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
tuple: Bars, particles, and counts displayed in the plot.
- prop(self, name)
- Return data about particles.
Parameter:
name: Key name for the particle data
Returns:
(list) particle data
- save(self, filename: str, compression='gzip')
- Saves the Graph object to a file.
Parameters:
filename (str): Path to the output file.
- top(self, prop='freq', n=1)
- Retrieve the top N particles based on a specified property.
Parameters:
prop (str): The property to sort by (default is 'freq').
n (int): The number of top particles to retrieve.
Returns:
np.ndarray: Array of the top N particles.
- update(self, particle, data, **kwargs)
- Adds particle to distribution. If particle already exists, update its data.
Parameter:
particle (Hashable): Particle to add
data (dict): Particle data
Class methods inherited from Distribution:
- compute_distribution(data: pandas.core.frame.DataFrame, score: structure_learning.scores.score.Score, graph_type='dag', blocklist: numpy.ndarray = None) -> Type[ForwardRef('D')]
- Compute a distribution from data and a scoring function.
Parameters:
data (pd.DataFrame): The dataset to compute the distribution from.
score (Score): The scoring function to evaluate particles.
graph_type (str): The type of graph to use ('dag' or 'cpdag').
Returns:
Distribution: The computed distribution.
- load(filename: str, compression='gzip')
- Loads a Graph object from a file.
Parameters:
filename (str): Path to the input file.
Returns:
Graph: Loaded Graph object.
- plot_multiple(dists: List[Type[ForwardRef('D')]], prop='freq', normalise=False, limit=-1, ax=None, labels=None)
- Plot multiple distributions on the same axes.
Parameters:
dists (List[Distribution]): List of distributions to plot.
prop (str): The property to plot (default is 'freq').
sort (bool): Whether to sort the particles by the property values.
normalise (bool): Whether to normalise the property values.
limit (int): The maximum number of particles to display (default is -1, which shows all).
ax (matplotlib.axes.Axes): The axes to plot on (default is None).
Returns:
list: List of bar containers for each distribution.
Readonly properties inherited from Distribution:
- logp
- Retrieve the log probabilities of all particles in the distribution.
Returns:
np.ndarray: Array of log probabilities.
Data descriptors inherited from Distribution:
- __dict__
- dictionary for instance variables
- __weakref__
- list of weak references to the object
| |