deadleaves.distributions#

Classes#

`DistSpec`	Distribution data class
`BaseDistribution`	Base class for custom distributions
`PowerLaw`	Distribution with density that follows the power law.
`Cosine`	Distribution with density that follows the cosine
`ExpCosine`	Distribution with cosine density and exponential peaks.
`Constant`	Distribution class which return a constant value.
`Image`	Distribution to sample images uniformly from an image data set.

Functions#

get_dist_kw(→ dict[str, DistSpec])

Return dictionary mapping keys to distribution classes and required parameters.

Module Contents#

class deadleaves.distributions.DistSpec#

Distribution data class

cls: Type[torch.distributions.distribution.Distribution]#: Distribution class

required: set[str]#: Parameters of the distribution

deadleaves.distributions.get_dist_kw() → dict[str, DistSpec]#: Return dictionary mapping keys to distribution classes and required parameters.

class deadleaves.distributions.BaseDistribution(*args, **kwargs)#

Bases: torch.distributions.distribution.Distribution

Base class for custom distributions

Raises:

NotImplementedError:

Class contains empty methods for

initialization (__init__)
probability density function (pdf)
inverse cumulative function (icdf)

_batch_shape: torch.Size#: The shape over which parameters are batched.

_validate_args()#

Validate input arguments.

Raises:: ValueError

abstractmethod pdf(x: torch.Tensor) → torch.Tensor#

Returns the probability density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

cdf(x: torch.Tensor) → torch.Tensor#

Returns (approximation of) the cumulative density function evaluated at x. If not implemented this method will compute the cdf as approximation via the pdf.

Args:

x (torch.Tensor):: Value(s) to evaluate.

abstractmethod icdf(p: torch.Tensor) → torch.Tensor#

Returns the inverse cumulative density function evaluated at p.

Args:

p (torch.Tensor):: Probability value(s) to evaluate.

ppf(p: torch.Tensor) → torch.Tensor#

Returns the percent point function, i.e. approximation of inverse cumulative function evaluated at p.

Args:: p (torch.Tensor): Probability value(s) to evaluate.

sample(n=1) → torch.Tensor#

Generates a sample from the distribution.

Args:: n (int, optional): Number of samples. Defaults to 1.

class deadleaves.distributions.PowerLaw(low: float, high: float, k: float = 3)#

Bases: BaseDistribution

Distribution with density that follows the power law.

Args:

low (float):: Minimal allowed value.
high (float):: Maximal allowed value.
k (float, optional):: Power law exponent. Defaults to 3.

property arg_constraints: dict#: Returns a dictionary from argument names to Constraint objects that should be satisfied by each argument of this distribution. Args that are not tensors need not appear in this dict.

__str__() → str#: Returns a readable string representation of the object.

property support#: Returns a Constraint object representing this distribution’s support.

pdf(x: torch.Tensor) → torch.Tensor#

Returns the probability density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

cdf(x: torch.Tensor) → torch.Tensor#

Returns (approximation of) the cumulative density function evaluated at x. If not implemented this method will compute the cdf as approximation via the pdf.

Args:

x (torch.Tensor):: Value(s) to evaluate.

icdf(p: torch.Tensor) → torch.Tensor#

Returns the inverse cumulative density function evaluated at p.

Args:

p (torch.Tensor):: Probability value(s) to evaluate.

class deadleaves.distributions.Cosine(amplitude: float = 0.5, frequency: int = 4)#

Bases: BaseDistribution

Distribution with density that follows the cosine

Args:

amplitude (float):: Amplitude of cosine. Value must be between 0.0 and 1.0. Defaults to 0.5.
frequency (int):: Frequency of cosine. Defaults to 4, i.e. peaks at the cardinals.

property support#: Returns a Constraint object representing this distribution’s support.

property arg_constraints#: Returns a dictionary from argument names to Constraint objects that should be satisfied by each argument of this distribution. Args that are not tensors need not appear in this dict.

__str__() → str#: Returns a readable string representation of the object.

pdf(x: torch.Tensor) → torch.Tensor#

Returns the probability density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

cdf(x: torch.Tensor) → torch.Tensor#

Returns (approximation of) the cumulative density function evaluated at x. If not implemented this method will compute the cdf as approximation via the pdf.

Args:

x (torch.Tensor):: Value(s) to evaluate.

class deadleaves.distributions.ExpCosine(frequency: int = 4, exponential_constant: float = 3)#

Bases: BaseDistribution

Distribution with cosine density and exponential peaks.

Args:

frequency (int):: Frequency of cosine. Defaults to 4, i.e. peaks at the cardinals.
exponential_constant (float):: Growth constant of exponential component. Larger values generate stronger peaks. Negative values invert the peaks. Defaults to 3.

property support#: Returns a Constraint object representing this distribution’s support.

property arg_constraints#: Returns a dictionary from argument names to Constraint objects that should be satisfied by each argument of this distribution. Args that are not tensors need not appear in this dict.

__str__() → str#: Returns a readable string representation of the object.

pdf(x: torch.Tensor) → torch.Tensor#

Returns the probability density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

class deadleaves.distributions.Constant(value: float)#

Bases: torch.distributions.distribution.Distribution

Distribution class which return a constant value.

Args:

value (float):: Constant value to be return in each sampling.

property arg_constraints#: Returns a dictionary from argument names to Constraint objects that should be satisfied by each argument of this distribution. Args that are not tensors need not appear in this dict.

property support#: Returns a Constraint object representing this distribution’s support.

_batch_shape: torch.Size#: The shape over which parameters are batched.

value: float#

__str__() → str#: Returns a readable string representation of the object.

pdf(x: torch.Tensor) → torch.Tensor#

Returns the probability density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

cdf(x: torch.Tensor) → torch.Tensor#

Returns the cumulative density function evaluated at x.

Args:

x (torch.Tensor):: Value(s) to evaluate.

icdf(p: torch.Tensor) → torch.Tensor#

Returns the inverse cumulative density function evaluated at p.

Args:

p (torch.Tensor):: Probability value(s) to evaluate.

sample(n=1) → torch.Tensor#

Generates a sample from the distribution.

Args:: n (int, optional): Number of samples. Defaults to 1.

class deadleaves.distributions.Image(dir: pathlib.Path | str)#

Bases: torch.distributions.distribution.Distribution

Distribution to sample images uniformly from an image data set.

Args:

dir (Path | str):: Path to image data set directory.

property arg_constraints#: Returns a dictionary from argument names to Constraint objects that should be satisfied by each argument of this distribution. Args that are not tensors need not appear in this dict.

property support#: Returns a Constraint object representing this distribution’s support.

_batch_shape: torch.Size#: The shape over which parameters are batched.

dir: pathlib.Path | str#: Path to image data set directory.

files: list[pathlib.Path]#: List of image files in the directory.

n_files: int#: Number of image files.

__str__() → str#: Returns a readable string representation of the object.

sample(n=1) → list[pathlib.Path]#

Draws a sample from the available images.

Args:: n (int, optional): Number of samples. Defaults to 1.