API Reference

This is the class and function reference of destructive-deep-learning (ddl).

ddl.autoregressive

Module for very simple autoregressive destructors.

Autoregressive classes

autoregressive.AutoregressiveDestructor([…])

Autoregressive destructor using densities that can compute conditionals.

ddl.base

Base destructors and destructor mixins.

Base classes

base.BaseDensityDestructor	Abstract destructor derived from an explicit underlying density.
base.BoundaryWarning	Warning that data is on the boundary of the required set.
base.CompositeDestructor([destructors, …])	Meta destructor composed of multiple destructors.
base.DestructorMixin	Mixin helper class to add universal destructor methods.
base.IdentityDestructor	Identity destructor/transform.
base.ScoreMixin	Mixin for score that returns mean of score_samples.
base.UniformDensity()	Uniform density estimator.

Base functions

base.create_implicit_density(fitted_destructor)	Create the implicit density associated with a fitted destructor.
base.create_inverse_canonical_destructor(…)	Create inverse destructor of a fitted canonical destructor.
base.create_inverse_transformer(…[, copy])	Create inverse transformer from fitted transformer.
base.get_implicit_density(\*args, \*\*kwargs)
base.get_inverse_canonical_destructor(…)
base.get_n_features(destructor[, …])	Get the number of features for a fitted destructor.

ddl.datasets

Simple module to generate toy datasets.

Datasets functions

datasets.make_toy_data(data_name[, …])

Make simple toy datasets.

ddl.deep

Deep destructors module.

Deep classes

deep.DeepDestructor([canonical_destructor, …])	Destructor formed by composing copies of some atomic destructor.
deep.DeepDestructorCV([…])	Deep destructor whose number of destructors/layers is determined by CV.

ddl.externals

Externals package init file.

ddl.externals.mlpack

Init for ddl.externals.mlpack package.

Mlpack classes

mlpack.MlpackDensityTreeEstimator([…])

Density tree estimator via mlpack (mlpack.org).

ddl.gaussian

Module for Gaussian density.

Gaussian classes

gaussian.GaussianDensity([covariance_type, …])

Gaussian density estimator with conditional and marginal methods.

ddl.independent

Module for independent densities and destructors.

Independent classes

independent.IndependentDensity([…])	Independent density estimator.
independent.IndependentDestructor([…])	Coordinate-wise destructor based on underlying independent density.
independent.IndependentInverseCdf	Independent inverse CDF transformer applied coordinate-wise.

ddl.linear

Module to handle linear projectors and destructors.

Linear classes

linear.BestLinearReconstructionDestructor([…])	Best linear reconstruction after a linear destructor.
linear.IdentityLinearEstimator()	Identity linear projection.
linear.LinearProjector([linear_estimator, …])	A linear projector based on an underlying linear estimator.
linear.RandomOrthogonalEstimator([…])	Random linear orthogonal estimator.

ddl.local

Module for local destructors such as adjacent-pixel-pair destructors.

Local classes

local.FeatureGroupsDestructor([…])	Destructor that transforms groups of features independently.
local.ImageFeaturePairs([image_shape, …])	Generate pairs of pixels based on image layout.
local.RandomFeaturePairs([random_state])	Random feature pairs estimator for use with FeatureGroupsDestructor.

ddl.mixture

Module for mixture densities and destructors.

Mixture classes

mixture.FirstFixedGaussianMixtureDensity([…])	Mixture density where one component is fixed as the standard normal.
mixture.GaussianMixtureDensity([…])	Gaussian mixture density that can be used with AutoregressiveDestructor.

ddl.tree

Module for tree densities and destructors.

Tree classes

tree.RandomTreeEstimator([min_samples_leaf, …])	Random tree estimator via ExtraTreeRegressor.
tree.TreeDensity([tree_estimator, get_tree, …])	Tree density estimator defined on the unit hypercube.
tree.TreeDestructor([tree_density])	Canonical tree destructor based on an underlying tree density.

ddl.univariate

Module for univariate densities (see also ddl.independent).

Univariate classes

univariate.HistogramUnivariateDensity([…])	Histogram univariate density estimator.
univariate.ScipyUnivariateDensity([…])	Density estimator via random variables defined in scipy.stats.

ddl.utils

Module for utility functions and classes.

Utils functions

utils.check_X_in_interval(X, interval)	Check if the input X lies in the specified interval.
utils.check_X_in_interval_decorator(func)	Decorate functions such as transform to check domain.
utils.check_domain(domain, n_features)	Check and return domain, broadcasting domain if necessary.
utils.get_domain_or_default(destructor[, warn])	Get the domain of the density or return DEFAULT_DOMAIN.
utils.get_support_or_default(density[, warn])	Get the support of the density or return DEFAULT_SUPPORT.
utils.has_method(est, method_name[, warn])	Check if an estimator has a method and possibly warn if not.
utils.make_finite(X)	Make the data matrix finite by replacing -infty and infty.
utils.make_interior(X, bounds[, eps])	Scale/shift data to fit in the open interval given by bounds.
utils.make_interior_probability(X[, eps])	Convert data to probability values in the open interval between 0 and 1.
utils.make_positive(X)	Make the data matrix positive by clipping to +epsilon if not positive.

ddl.validation

Module for validating destructor interface and testing properties.

Validation functions

validation.check_density(density[, random_state])	Check that an estimator implements density methods correctly.
validation.check_destructor(destructor[, …])	Check for the required interface and properties of a destructor.