alibi.explainers.backends.tensorflow.cfrl_base

This module contains utility functions for the Counterfactual with Reinforcement Learning base class, :py:class:alibi.explainers.cfrl_base, for the Tensorflow backend.

Constants

TYPE_CHECKING

TYPE_CHECKING: bool = False

bool(x) -> bool

Returns True when the argument x is true, False otherwise. The builtins True and False are the only two instances of the class bool. The class bool is a subclass of the class int, and cannot be subclassed.

TfCounterfactualRLDataset

Inherits from: CounterfactualRLDataset, ABC, PyDataset

Tensorflow backend datasets.

Constructor

TfCounterfactualRLDataset(self, X: numpy.ndarray, preprocessor: Callable, predictor: Callable, conditional_func: Callable, batch_size: int, shuffle: bool = True) -> None

Name	Type	Default	Description
X	numpy.ndarray		Array of input instances. The input should NOT be preprocessed as it will be preprocessed when calling the preprocessor function.
preprocessor	Callable		Preprocessor function. This function correspond to the preprocessing steps applied to the encoder/auto-encoder model.
predictor	Callable		Prediction function. The classifier function should expect the input in the original format and preprocess it internally in the predictor if necessary.
conditional_func	Callable		Conditional function generator. Given an pre-processed input array, the functions generates a conditional array.
batch_size	int		Dimension of the batch used during training. The same batch size is used to infer the classification labels of the input dataset.
shuffle	bool	True	Whether to shuffle the dataset each epoch. True by default.

Methods

on_epoch_end

on_epoch_end() -> None

This method is called every epoch and performs dataset shuffling.

Returns

• Type: None

Functions

add_noise

add_noise(Z_cf: Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray], noise: NormalActionNoise, act_low: float, act_high: float, step: int, exploration_steps: int, kwargs) -> tensorflow.python.framework.tensor.Tensor

Add noise to the counterfactual embedding.

Name	Type	Default	Description
Z_cf	Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray]		Counterfactual embedding.
noise	NormalActionNoise		Noise generator object.
act_low	float		Noise lower bound.
act_high	float		Noise upper bound.
step	int		Training step.
exploration_steps	int		Number of exploration steps. For the first exploration_steps, the noised counterfactual embedding is sampled uniformly at random.
**kwargs			Other arguments. Not used.

Returns

• Type: tensorflow.python.framework.tensor.Tensor

consistency_loss

consistency_loss(Z_cf_pred: tensorflow.python.framework.tensor.Tensor, Z_cf_tgt: tensorflow.python.framework.tensor.Tensor)

Default 0 consistency loss.

Name	Type	Default	Description
Z_cf_pred	tensorflow.python.framework.tensor.Tensor		Counterfactual embedding prediction.
Z_cf_tgt	tensorflow.python.framework.tensor.Tensor		Counterfactual embedding target.

data_generator

data_generator(X: numpy.ndarray, encoder_preprocessor: Callable, predictor: Callable, conditional_func: Callable, batch_size: int, shuffle: bool = True, kwargs)

Constructs a tensorflow data generator.

Name	Type	Default	Description
X	numpy.ndarray		Array of input instances. The input should NOT be preprocessed as it will be preprocessed when calling the preprocessor function.
encoder_preprocessor	Callable		Preprocessor function. This function correspond to the preprocessing steps applied to the encoder/auto-encoder model.
predictor	Callable		Prediction function. The classifier function should expect the input in the original format and preprocess it internally in the predictor if necessary.
conditional_func	Callable		Conditional function generator. Given an preprocessed input array, the functions generates a conditional array.
batch_size	int		Dimension of the batch used during training. The same batch size is used to infer the classification labels of the input dataset.
shuffle	bool	True	Whether to shuffle the dataset each epoch. True by default.
**kwargs			Other arguments. Not used.

decode

decode(Z: Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray], decoder: keras.src.models.model.Model, kwargs)

Decodes an embedding tensor.

Name	Type	Default	Description
Z	Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray]		Embedding tensor to be decoded.
decoder	keras.src.models.model.Model		Pretrained decoder network.
**kwargs			Other arguments. Not used.

encode

encode(X: Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray], encoder: keras.src.models.model.Model, kwargs) -> tensorflow.python.framework.tensor.Tensor

Encodes the input tensor.

Name	Type	Default	Description
X	Union[tensorflow.python.framework.tensor.Tensor, numpy.ndarray]		Input to be encoded.
encoder	keras.src.models.model.Model		Pretrained encoder network.
**kwargs			Other arguments. Not used.

Returns

• Type: tensorflow.python.framework.tensor.Tensor

generate_cf

generate_cf(Z: Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor], Y_m: Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor], Y_t: Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor], C: Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor, None], actor: keras.src.models.model.Model, kwargs) -> tensorflow.python.framework.tensor.Tensor

Generates counterfactual embedding.

Name	Type	Default	Description
Z	Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor]		Input embedding tensor.
Y_m	Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor]		Input classification label.
Y_t	Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor]		Target counterfactual classification label.
C	Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor, None]		Conditional tensor.
actor	keras.src.models.model.Model		Actor network. The model generates the counterfactual embedding.
**kwargs			Other arguments. Not used.

Returns

• Type: tensorflow.python.framework.tensor.Tensor

get_actor

get_actor(hidden_dim: int, output_dim: int) -> keras.src.layers.layer.Layer

Constructs the actor network.

Name	Type	Default	Description
hidden_dim	int		Actor's hidden dimension
output_dim	int		Actor's output dimension.

Returns

• Type: keras.src.layers.layer.Layer

get_critic

get_critic(hidden_dim: int) -> keras.src.layers.layer.Layer

Constructs the critic network.

Name	Type	Default	Description
hidden_dim	int		Critic's hidden dimension.

Returns

• Type: keras.src.layers.layer.Layer

get_optimizer

get_optimizer(model: Optional[keras.src.layers.layer.Layer] = None, lr: float = 0.001) -> keras.src.optimizers.optimizer.Optimizer

Constructs default Adam optimizer.

Name	Type	Default	Description
model	Optional[keras.src.layers.layer.Layer]	None	Model to get the optimizer for. Not required for tensorflow backend.
lr	float	0.001	Learning rate.

Returns

• Type: keras.src.optimizers.optimizer.Optimizer

initialize_actor_critic

initialize_actor_critic(actor, critic, Z, Z_cf_tilde, Y_m, Y_t, C, kwargs)

Initialize actor and critic layers by passing a dummy zero tensor.

Name	Type	Default	Description
actor			Actor model.
critic			Critic model.
Z			Input embedding.
Z_cf_tilde			Noised counterfactual embedding.
Y_m			Input classification label.
Y_t			Target counterfactual classification label.
C			Conditional tensor.
**kwargs			Other arguments. Not used.

initialize_optimizer

initialize_optimizer(optimizer: keras.src.optimizers.optimizer.Optimizer, model: keras.src.models.model.Model) -> None

Initializes an optimizer given a model.

Name	Type	Default	Description
optimizer	keras.src.optimizers.optimizer.Optimizer		Optimizer to be initialized.
model	keras.src.models.model.Model		Model to be optimized

Returns

• Type: None

initialize_optimizers

initialize_optimizers(optimizer_actor, optimizer_critic, actor, critic, kwargs) -> None

Initializes the actor and critic optimizers.

Name	Type	Default	Description
optimizer_actor			Actor optimizer to be initialized.
optimizer_critic			Critic optimizer to be initialized.
actor			Actor model to be optimized.
critic			Critic model to be optimized.
**kwargs			Other arguments. Not used.

Returns

• Type: None

load_model

load_model(path: Union[str, os.PathLike]) -> keras.src.models.model.Model

Loads a model and its optimizer.

Name	Type	Default	Description
path	Union[str, os.PathLike]		Path to the loading location.

Returns

• Type: keras.src.models.model.Model

save_model

save_model(path: Union[str, os.PathLike], model: keras.src.layers.layer.Layer) -> None

Saves a model and its optimizer.

Name	Type	Default	Description
path	Union[str, os.PathLike]		Path to the saving location.
model	keras.src.layers.layer.Layer		Model to be saved.

Returns

• Type: None

set_seed

set_seed(seed: int = 13)

Sets a seed to ensure reproducibility. Does NOT ensure reproducibility.

Name	Type	Default	Description
seed	int	13	seed to be set

sparsity_loss

sparsity_loss(X_hat_cf: tensorflow.python.framework.tensor.Tensor, X: tensorflow.python.framework.tensor.Tensor) -> Dict[str, tensorflow.python.framework.tensor.Tensor]

Default L1 sparsity loss.

Name	Type	Default	Description
X_hat_cf	tensorflow.python.framework.tensor.Tensor		Auto-encoder counterfactual reconstruction.
X	tensorflow.python.framework.tensor.Tensor		Input instance.

Returns

• Type: Dict[str, tensorflow.python.framework.tensor.Tensor]

to_numpy

to_numpy(X: Union[List[Any], numpy.ndarray, tensorflow.python.framework.tensor.Tensor, None]) -> Union[List[Any], numpy.ndarray, None]

Converts given tensor to numpy array.

Name	Type	Default	Description
X	Union[List[Any], numpy.ndarray, tensorflow.python.framework.tensor.Tensor, None]		Input tensor to be converted to numpy array.

Returns

• Type: Union[List[Any], numpy.ndarray, None]

to_tensor

to_tensor(X: Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor], kwargs) -> Optional[tensorflow.python.framework.tensor.Tensor]

Converts tensor to tf.Tensor.

Name	Type	Default	Description
X	Union[numpy.ndarray, tensorflow.python.framework.tensor.Tensor]		Input array/tensor to be converted.
**kwargs			Other arguments. Not used.

Returns

• Type: Optional[tensorflow.python.framework.tensor.Tensor]