alibi_detect.utils.visualize

Functions

plot_feature_outlier_image

plot_feature_outlier_image(od_preds: Dict, X: numpy.ndarray, X_recon: Optional[numpy.ndarray] = None, instance_ids: Optional[list] = None, max_instances: int = 5, outliers_only: bool = False, n_channels: int = 3, figsize: tuple = (20, 20)) -> None

Plot feature (pixel) wise outlier scores for images.

Name	Type	Default	Description
od_preds	Dict		Output of an outlier detector's prediction.
X	numpy.ndarray		Batch of instances to apply outlier detection to.
X_recon	Optional[numpy.ndarray]	None	Reconstructed instances of X.
instance_ids	Optional[list]	None	List with indices of instances to display.
max_instances	int	5	Maximum number of instances to display.
outliers_only	bool	False	Whether to only show outliers or not.
n_channels	int	3	Number of channels of the images.
figsize	tuple	(20, 20)	Tuple for the figure size.

Returns

• Type: None

plot_feature_outlier_tabular

plot_feature_outlier_tabular(od_preds: Dict, X: numpy.ndarray, X_recon: Optional[numpy.ndarray] = None, threshold: Optional[float] = None, instance_ids: Optional[list] = None, max_instances: int = 5, top_n: int = 1000000000000, outliers_only: bool = False, feature_names: Optional[list] = None, width: float = 0.2, figsize: tuple = (20, 10)) -> None

Plot feature wise outlier scores for tabular data.

Name	Type	Default	Description
od_preds	Dict		Output of an outlier detector's prediction.
X	numpy.ndarray		Batch of instances to apply outlier detection to.
X_recon	Optional[numpy.ndarray]	None	Reconstructed instances of X.
threshold	Optional[float]	None	Threshold used for outlier score to determine outliers.
instance_ids	Optional[list]	None	List with indices of instances to display.
max_instances	int	5	Maximum number of instances to display.
top_n	int	1000000000000	Maixmum number of features to display, ordered by outlier score.
outliers_only	bool	False	Whether to only show outliers or not.
feature_names	Optional[list]	None	List with feature names.
width	float	0.2	Column width for bar charts.
figsize	tuple	(20, 10)	Tuple for the figure size.

Returns

• Type: None

plot_feature_outlier_ts

plot_feature_outlier_ts(od_preds: Dict, X: numpy.ndarray, threshold: Union[float, int, list, numpy.ndarray], window: Optional[tuple] = None, t: Optional[numpy.ndarray] = None, X_orig: Optional[numpy.ndarray] = None, width: float = 0.2, figsize: tuple = (20, 8), ylim: tuple = (None, None)) -> None

Plot feature wise outlier scores for time series data.

Name	Type	Default	Description
od_preds	Dict		Output of an outlier detector's prediction.
X	numpy.ndarray		Time series to apply outlier detection to.
threshold	Union[float, int, list, numpy.ndarray]		Threshold used to classify outliers or adversarial instances.
window	Optional[tuple]	None	Start and end timestep to plot.
t	Optional[numpy.ndarray]	None	Timesteps.
X_orig	Optional[numpy.ndarray]	None	Optional original time series without outliers.
width	float	0.2	Column width for bar charts.
figsize	tuple	(20, 8)	Tuple for the figure size.
ylim	tuple	(None, None)	Min and max y-axis values for the outlier scores.

Returns

• Type: None

plot_instance_score

plot_instance_score(preds: Dict, target: numpy.ndarray, labels: numpy.ndarray, threshold: float, ylim: tuple = (None, None)) -> None

Scatter plot of a batch of outlier or adversarial scores compared to the threshold.

Name	Type	Default	Description
preds	Dict		Dictionary returned by predictions of an outlier or adversarial detector.
target	numpy.ndarray		Ground truth.
labels	numpy.ndarray		List with names of classification labels.
threshold	float		Threshold used to classify outliers or adversarial instances.
ylim	tuple	(None, None)	Min and max y-axis values.

Returns

• Type: None

plot_roc

plot_roc(roc_data: Dict[str, Dict[str, numpy.ndarray]], figsize: tuple = (10, 5)) -> None

Plot ROC curve.

Name	Type	Default	Description
roc_data	Dict[str, Dict[str, numpy.ndarray]]		Dictionary with as key the label to show in the legend and as value another dictionary with as keys scores and labels with respectively the outlier scores and outlier labels.
figsize	tuple	(10, 5)	Figure size.

Returns

• Type: None