mojolearn.EllipticEnvelope#

class mojolearn.EllipticEnvelope(*, store_precision=True, assume_centered=False, support_fraction=None, contamination=0.1, random_state=None, numeric_mode=None)[source]#

Bases: MinCovDet

Outlier detection for Gaussian-distributed data.

Follows scikit-learn’s EllipticEnvelope: fits a MinCovDet and sets offset_ at the contamination percentile of the negated training distances, so that about that fraction of training points is labeled an outlier.

Parameters:

  • store_precision (bool, default True) – Store precision_.

  • assume_centered (bool, default False) – Treat the data as centered.

  • support_fraction (float or None, default None) – Fraction of points in the raw support; see MinCovDet.

  • contamination (float, default 0.1) – Expected fraction of outliers, in (0, 0.5].

  • random_state (int or None, default None) – Seed for the random subsets; None means 0.

  • numeric_mode ({“identical”, “fast”} or None, default None) – Numeric mode for this estimator. None uses the package default ("identical" unless changed). "identical" gives the same bits on NVIDIA and AMD GPUs; "fast" applies on Apple GPUs.

Variables:

  • offset_ (float) – Threshold on the decision function.

  • dist_ (location_, covariance_, precision_, support_,) – As in MinCovDet.

Raises:

ValueError – If contamination is not in (0, 0.5].

Examples

>>> import numpy as np, mojolearn
>>> X = np.random.default_rng(0).normal(size=(500, 3)).astype(np.float32)
>>> labels = mojolearn.EllipticEnvelope(contamination=0.05, random_state=0).fit(X).predict(X)
__init__(*, store_precision=True, assume_centered=False, support_fraction=None, contamination=0.1, random_state=None, numeric_mode=None)[source]#

Methods

decision_function(X)

Return score_samples(X) - offset_; negative means outlier.

fit(X[, y])

Fit the robust model and set offset_.

fit_predict(X[, y])

Fit the model, then return predict(X).

fit_transform(X[, y])

Fit to X, then return transform(X).

get_params([deep])

Return this estimator's constructor parameters.

get_precision()

Return the precision matrix (pseudo-inverse of covariance_).

mahalanobis(X)

Return squared Mahalanobis distances under the fitted model.

predict(X)

Label each sample 1 (inlier) or -1 (outlier).

score(X, y[, sample_weight])

Return the accuracy of predict(X) against labels y.

score_samples(X)

Return the negated squared Mahalanobis distance of each sample.

set_params(**params)

Set constructor parameters by keyword.
fit(X, y=None)[source]#

Fit the robust model and set offset_.

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • y (ignored)

Returns:

self

score_samples(X)[source]#

Return the negated squared Mahalanobis distance of each sample.

Parameters:

X (array-like of shape (n_samples, n_features))

Returns:

ndarray of shape (n_samples,), float32

decision_function(X)[source]#

Return score_samples(X) - offset_; negative means outlier.

Parameters:

X (array-like of shape (n_samples, n_features))

Returns:

ndarray of shape (n_samples,), float32

predict(X)[source]#

Label each sample 1 (inlier) or -1 (outlier).

Parameters:

X (array-like of shape (n_samples, n_features))

Returns:

ndarray of shape (n_samples,), int32

fit_predict(X, y=None)[source]#

Fit the model, then return predict(X).

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • y (ignored)

Returns:

ndarray of shape (n_samples,), int32

score(X, y, sample_weight=None)[source]#

Return the accuracy of predict(X) against labels y.

As scikit-learn’s EllipticEnvelope.score: the (weighted) fraction of exact label matches.

Parameters:

  • X (array-like of shape (n_samples, n_features))

  • y (array-like of shape (n_samples,)) – True labels (1 inlier, -1 outlier).

  • sample_weight (array-like of shape (n_samples,) or None, default None)

Returns:

float

Raises: