autoprognosis.plugins.explainers.plugin_invase module

class INVASEPlugin(estimator: Any, X: pandas.core.frame.DataFrame, y: pandas.core.frame.DataFrame, time_to_event: Optional[pandas.core.frame.DataFrame] = None, eval_times: Optional[List] = None, feature_names: Optional[List] = None, n_epoch: int = 10000, n_epoch_inner: int = 2, n_folds: int = 5, task_type: str = 'classification', samples: int = 2000, prefit: bool = False, random_state: int = 0)

Bases: autoprognosis.plugins.explainers.base.ExplainerPlugin

Interpretability plugin based on the INVASE algorithm.

Parameters

  • estimator – model. The model to explain.

  • X – dataframe. Training set

  • y – dataframe. Training labels

  • time_to_event – dataframe. Used for risk estimation tasks.

  • eval_times – list. Used for risk estimation tasks.

  • n_epoch – int. training epochs

  • task_type – str. classification or risk_estimation

  • samples – int. Number of samples to use.

  • prefit – bool. If true, the estimator won’t be trained.

Example

>>> import pandas as pd
>>> from sklearn.datasets import load_iris
>>> from sklearn.model_selection import train_test_split
>>>from autoprognosis.plugins.explainers import Explainers
>>> from autoprognosis.plugins.prediction.classifiers import Classifiers
>>>
>>> X, y = load_iris(return_X_y=True)
>>>
>>> X = pd.DataFrame(X)
>>> y = pd.Series(y)
>>>
>>> X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
>>> model = Classifiers().get("logistic_regression")
>>>
>>> explainer = Explainers().get(
>>>     "invase",
>>>     model,
>>>     X_train,
>>>     y_train,
>>>     task_type="classification",
>>> )
>>>
>>> explainer.explain(X_test)
explain(X: pandas.core.frame.DataFrame) → numpy.ndarray
static name() → str
plot(values: pandas.core.frame.DataFrame) → None
static pretty_name() → str
static type() → str
class Masking(*args: Any, **kwargs: Any)

Bases: torch.nn.Module

forward(tensors: List[torch.Tensor]) → torch.Tensor
bitmask_intervals(n: int, low: int, high: int) → Generator
bitmasks(n: int, m: int) → Generator
class invaseBase(estimator: Any, X: numpy.ndarray, n_epoch: int = 10000, n_epoch_inner: int = 1, patience: int = 5, min_epochs: int = 100, n_epoch_print: int = 50, batch_size: int = 300, learning_rate: float = 0.001, penalty_l2: float = 0.001, feature_names: List = [])

Bases: object

abstract explain(X: numpy.ndarray, *args: Any, **kwargs: Any) → numpy.ndarray
class invaseCV(estimator: Any, X: numpy.ndarray, critic_latent_dim: int = 200, n_epoch: int = 10000, n_epoch_inner: int = 2, patience: int = 5, min_epochs: int = 100, n_epoch_print: int = 50, n_folds: int = 5, seed: int = 42, feature_names: List = [])

Bases: object

explain(x: numpy.ndarray) → numpy.ndarray
class invaseClassifier(estimator: Any, X: numpy.ndarray, critic_latent_dim: int = 200, n_epoch: int = 10000, n_epoch_inner: int = 2, patience: int = 5, min_epochs: int = 100, n_epoch_print: int = 50, batch_size: int = 300, learning_rate: float = 0.001, penalty_l2: float = 0.001, feature_names: List = [])

Bases: autoprognosis.plugins.explainers.plugin_invase.invaseBase

explain(X: numpy.ndarray, *args: Any, **kwargs: Any) → numpy.ndarray
class invaseRiskEstimation(estimator: Any, X: numpy.ndarray, eval_times: List, critic_latent_dim: int = 200, n_epoch: int = 10000, n_epoch_inner: int = 2, patience: int = 5, min_epochs: int = 100, n_epoch_print: int = 10, batch_size: int = 500, learning_rate: float = 0.001, penalty_l2: float = 0.001, samples: int = 20000, feature_names: List = [])

Bases: autoprognosis.plugins.explainers.plugin_invase.invaseBase

explain(X: numpy.ndarray, *args: Any, **kwargs: Any) → numpy.ndarray
plugin

alias of autoprognosis.plugins.explainers.plugin_invase.INVASEPlugin

sample(X: numpy.ndarray, nsamples: int = 100, random_state: int = 0) → numpy.ndarray