alibi_detect.utils.mapping

Functions

num2ord

num2ord(data: numpy.ndarray, dist: dict) -> numpy.ndarray

Transform numerical values into categories using the map calculated under the fit method.

Name	Type	Default	Description
data	numpy.ndarray		Numpy array with the numerical data.
dist	dict		Dict with as keys the categorical variables and as values the numerical value for each category.

Returns

• Type: numpy.ndarray

ohe2ord

ohe2ord(X_ohe: numpy.ndarray, cat_vars_ohe: dict) -> Tuple[numpy.ndarray, dict]

Convert one-hot encoded variables to ordinal encodings.

Name	Type	Default	Description
X_ohe	numpy.ndarray		Data with mixture of one-hot encoded and numerical variables.
cat_vars_ohe	dict		Dict with as keys the first column index for each one-hot encoded categorical variable and as values the number of categories per categorical variable.

Returns

• Type: Tuple[numpy.ndarray, dict]

ohe2ord_shape

ohe2ord_shape(shape: tuple, cat_vars: Optional[dict] = None, is_ohe: bool = False) -> tuple

Infer shape of instance if the categorical variables have ordinal instead of on-hot encoding.

Name	Type	Default	Description
shape	tuple		Instance shape, starting with batch dimension.
cat_vars	Optional[dict]	None	Dict with as keys the categorical columns and as values the number of categories per categorical variable.
is_ohe	bool	False	Whether instance is OHE.

Returns

• Type: tuple

ord2num

ord2num(data: numpy.ndarray, dist: dict) -> numpy.ndarray

Transform categorical into numerical values using a mapping.

Name	Type	Default	Description
data	numpy.ndarray		Numpy array with the categorical data.
dist	dict		Dict with as keys the categorical variables and as values the numerical value for each category.

Returns

• Type: numpy.ndarray

ord2ohe

ord2ohe(X_ord: numpy.ndarray, cat_vars_ord: dict) -> Tuple[numpy.ndarray, dict]

Convert ordinal to one-hot encoded variables.

Name	Type	Default	Description
X_ord	numpy.ndarray		Data with mixture of ordinal encoded and numerical variables.
cat_vars_ord	dict		Dict with as keys the categorical columns and as values the number of categories per categorical variable.

Returns

• Type: Tuple[numpy.ndarray, dict]