lczerolens.concepts#

Class for concept-based XAI methods.

Classes#

`Concept`	Class for concept-based XAI methods.
`BinaryConcept`	Class for binary concept-based XAI methods.
`NullConcept`	Class for binary concept-based XAI methods.
`OrBinaryConcept`	Class for binary concept-based XAI methods.
`AndBinaryConcept`	Class for binary concept-based XAI methods.
`MulticlassConcept`	Class for multiclass concept-based XAI methods.
`ContinuousConcept`	Class for continuous concept-based XAI methods.
`HasPiece`	Class for material concept-based XAI methods.
`HasMaterialAdvantage`	Class for material concept-based XAI methods.
`BestLegalMove`	Class for move concept-based XAI methods.
`PieceBestLegalMove`	Class for move concept-based XAI methods.
`HasThreat`	Class for material concept-based XAI methods.
`HasMateThreat`	Class for material concept-based XAI methods.

Module Contents#

class lczerolens.concepts.Concept[source]#

Bases: abc.ABC

Class for concept-based XAI methods.

abstract compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: Any

static compute_metrics(predictions, labels)[source]#

Abstractmethod:

Compute the metrics for a given model and input.

static get_dataset_feature()[source]#

Abstractmethod:

Returns the feature for the dataset.

class lczerolens.concepts.BinaryConcept[source]#

Bases: Concept

Class for binary concept-based XAI methods.

static compute_metrics(predictions, labels)[source]#: Compute the metrics for a given model and input.

static get_dataset_feature()[source]#: Returns the feature for the dataset.

class lczerolens.concepts.NullConcept[source]#

Bases: BinaryConcept

Class for binary concept-based XAI methods.

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: Any

class lczerolens.concepts.OrBinaryConcept(*concepts)[source]#

Bases: BinaryConcept

Class for binary concept-based XAI methods.

Parameters:: concepts (BinaryConcept)

concepts = ()[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: Any

class lczerolens.concepts.AndBinaryConcept(*concepts)[source]#

Bases: BinaryConcept

Class for binary concept-based XAI methods.

Parameters:: concepts (BinaryConcept)

concepts = ()[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: Any

class lczerolens.concepts.MulticlassConcept[source]#

Bases: Concept

Class for multiclass concept-based XAI methods.

static compute_metrics(predictions, labels)[source]#: Compute the metrics for a given model and input.

static get_dataset_feature()[source]#: Returns the feature for the dataset.

class lczerolens.concepts.ContinuousConcept[source]#

Bases: Concept

Class for continuous concept-based XAI methods.

static compute_metrics(predictions, labels)[source]#: Compute the metrics for a given model and input.

static get_dataset_feature()[source]#: Returns the feature for the dataset.

class lczerolens.concepts.HasPiece(piece, relative=True)[source]#

Bases: BinaryConcept

Class for material concept-based XAI methods.

Parameters:

piece (str)
relative (bool)

piece[source]#

relative = True[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: int

class lczerolens.concepts.HasMaterialAdvantage(relative=True)[source]#

Bases: BinaryConcept

Class for material concept-based XAI methods.

Parameters:: relative (bool)

piece_values[source]#

The piece values.

Type:: Dict[int, int]

piece_values[source]#

relative = True[source]#

compute_label(board, piece_values=None)[source]#

Compute the label for a given model and input.

Parameters:

board (lczerolens.board.LczeroBoard)
piece_values (Optional[Dict[int, int]])

Return type:

int

class lczerolens.concepts.BestLegalMove(model)[source]#

Bases: MulticlassConcept

Class for move concept-based XAI methods.

Parameters:: model (lczerolens.model.LczeroModel)

policy_flow[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: int

class lczerolens.concepts.PieceBestLegalMove(model, piece)[source]#

Bases: BinaryConcept

Class for move concept-based XAI methods.

Parameters:

model (lczerolens.model.LczeroModel)
piece (str)

policy_flow[source]#

piece[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: int

class lczerolens.concepts.HasThreat(piece, relative=True)[source]#

Bases: BinaryConcept

Class for material concept-based XAI methods.

Parameters:

piece (str)
relative (bool)

piece[source]#

relative = True[source]#

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: int

class lczerolens.concepts.HasMateThreat[source]#

Bases: BinaryConcept

Class for material concept-based XAI methods.

compute_label(board)[source]#

Compute the label for a given model and input.

Parameters:: board (lczerolens.board.LczeroBoard)
Return type:: int