openasce.attribution package

class openasce.attribution.Attribution(*, threshold: float, max_step: int = 2, top_num: int = None)[source]

Bases: Runtime

Attribution Class

Attributes:

Constructor

Argument:

threshold: the score threshold max_step: the maximal step. For the attribution based on causal graph, that is the maximal node number. top_num: the accepted number of best options in each step, which is used in greedy attribution.

__doc__ = 'Attribution Class\n\n    Attributes:\n\n    '
__init__(*, threshold: float, max_step: int = 2, top_num: int = None) → None[source]

Constructor

Argument:

threshold: the score threshold max_step: the maximal step. For the attribution based on causal graph, that is the maximal node number. top_num: the accepted number of best options in each step, which is used in greedy attribution.

__module__ = 'openasce.attribution.attribution_model'
attribute(*, X: Iterable[ndarray], Y: Iterable[ndarray] = None, T: Iterable[ndarray] = None, **kwargs) → None[source]

Feed the sample data to attribute.

Parameters

  • X – Features of the samples.

  • Y – Ignore for now and keep for future

  • T – Ignore for now and keep for future

  • kwargs – {‘treat_value’: treat_value}, maximization when treat_value

Returns

None

property column_names

All nodes’ name. Note: should include the treatment node and label node.

get_result()[source]

Get the result

Returns

The attribution result.

property inferencer: InferenceModel

The inference object used to estimate the effect

property label_name
property label_value
property treatment_name

Submodules

openasce.attribution.attribution_model module

class openasce.attribution.attribution_model.Attribution(*, threshold: float, max_step: int = 2, top_num: int = None)[source]

Bases: Runtime

Attribution Class

Attributes:

Constructor

Argument:

threshold: the score threshold max_step: the maximal step. For the attribution based on causal graph, that is the maximal node number. top_num: the accepted number of best options in each step, which is used in greedy attribution.

__annotations__ = {}
__doc__ = 'Attribution Class\n\n    Attributes:\n\n    '
__init__(*, threshold: float, max_step: int = 2, top_num: int = None) → None[source]

Constructor

Argument:

threshold: the score threshold max_step: the maximal step. For the attribution based on causal graph, that is the maximal node number. top_num: the accepted number of best options in each step, which is used in greedy attribution.

__module__ = 'openasce.attribution.attribution_model'
attribute(*, X: Iterable[ndarray], Y: Iterable[ndarray] = None, T: Iterable[ndarray] = None, **kwargs) → None[source]

Feed the sample data to attribute.

Parameters

  • X – Features of the samples.

  • Y – Ignore for now and keep for future

  • T – Ignore for now and keep for future

  • kwargs – {‘treat_value’: treat_value}, maximization when treat_value

Returns

None

property column_names

All nodes’ name. Note: should include the treatment node and label node.

get_result()[source]

Get the result

Returns

The attribution result.

property inferencer: InferenceModel

The inference object used to estimate the effect

property label_name
property label_value
property treatment_name

openasce.attribution.attribution_test module