utils.imput package

Submodules

utils.imput.imputer module

exception utils.imput.imputer.ImputerException(message='')

Bases: VL3DException

Author:

Alberto M. Esmoris Pena

Class for exceptions related to imputation components. See VL3DException.

__init__(message='')
class utils.imput.imputer.Imputer(**kwargs)

Bases: object

Author:

Alberto M. Esmoris Pena

Class for imputation operations.

Variables:

  • target_val (str or int or float) – The target value, i.e., features that match this value will be imputed.

  • fnames (list or tuple) – The names of the features to be imputed (by default).

  • impute_coordinates (bool) – Whether to consider the point-wise coordinates for the imputation (True) or not (False, default).

  • impute_references (bool) – Whether to consider the point-wise references for the imputation (True) or not (False, default).

static extract_imputer_args(spec)

Extract the arguments to initialize/instantiate an Imputer from a key-word specification.

Parameters:

spec – The key-word specification containing the arguments.

Returns:

The arguments to initialize/instantiate an Imputer.

__init__(**kwargs)

Initialize/instantiate an Imputer.

Parameters:

kwargs – The attributes for the Imputer.

abstractmethod impute(F, y=None)

The fundamental imputation logic defining the imputer.

Parameters:

  • F – The input matrix of features to be imputed.

  • y – The input vector of classes as an optional argument. When given, the imputation in F will be propagated to y if necessary. This imputation is not often needed, but some strategies demand it to keep the consistency between features and expected classes. For example, when the imputation strategy consists of removing points with NaN values, the corresponding component from the vector of classes must also be removed.

Returns:

The matrix of features and the vector of classes after imputation. If the vector of classes is not given (i.e., is None), then only imputed F will be returned.

Return type:

np.ndarray or tuple

impute_pcloud(pcloud, fnames=None)

Apply the impute method to a point cloud.

See imputer.Imputer.impute().

Parameters:

  • pcloud (PointCloud) – The point cloud to be imputed.

  • fnames (list or tuple) – The list of features to be imputed. If None, it will be taken from the internal fnames of the imputer. If those are None too, then an exception will raise.

Returns:

The updated point cloud after the imputations.

Return type:

PointCloud

find_fnames(fnames=None)

Find the feature names. First, given ones will be considered. If they are not given, then member feature names will be considered if available. Otherwise, an exception will be thrown.

Parameters:

fnames – The list of features to be imputed. If None, it will be taken from the memember feature names of the imputer. If those are not available, then an exception will be thrown.

Returns:

The found feature names.

Return type:

list of str

extract_pcloud_matrix(pcloud, fnames)

Extract values from the point cloud to build a matrix representing it.

Parameters:

fnames – The names of the features that must be considered.

Returns:

The matrix representing the point cloud.

Return type:

np.ndarray

utils.imput.removal_imputer module

class utils.imput.removal_imputer.RemovalImputer(**kwargs)

Bases: Imputer

Author:

Alberto M. Esmoris Pena

Class to remove missing values.

__init__(**kwargs)

Initialize/instantiate a RemovalImputer

Parameters:

kwargs – The attributes for the RemovalImputer

impute(F, y=None)

The fundamental imputation logic defining the removal imputer.

See Imputer and imputer.Imputer.impute().

impute_pcloud(pcloud, fnames=None)

Overwrite the logic of Imputer.impute_pcloud() because RemovalImputer might need to remove points from the point cloud.

See Imputer.imput_pcloud().

find_target_mask(F)

Obtain a boolean mask that specifies whether a given point matches the target value (True) or not (False).

Parameters:

F – The matrix of point-wise features representing the point cloud.

Returns:

The

utils.imput.univariate_imputer module

class utils.imput.univariate_imputer.UnivariateImputer(**kwargs)

Bases: Imputer

Author:

Alberto M. Esmoris Pena

Class to compute univariate imputations.

static extract_imputer_args(spec)

Extract the arguments to initialize/instantiate an UnivariateImputer from a key-word specification.

Parameters:

spec – The key-word specification containing the arguments.

Returns:

The arguments to initialize/instantiate an UnivariateImputer.

__init__(**kwargs)

Initialize/instantiate a UnivariateImputer

Parameters:

kwargs – The attributes for the UnivariateImputer

impute(F, y=None)

The fundamental imputation logic defining the univariate imputer See Imputer

In this case, since imputation will not remove points, the y argument will be ignored, no matter what. However, in case y is given as not None, the return will be (imputed F, y) for compatibility and fluent programming. If y is None, only imputed F will be return.

Module contents

author:

Alberto M. Esmoris Pena

The imputation package contains the logic to impute missing values from point clouds and potential complementary data sources.