detectors.aggregations package

Submodules

detectors.aggregations.anomaly module

class detectors.aggregations.anomaly.AnomalyBaseAggregation(method_class, **kwargs)

Bases: object

fit(stack: Tensor, *args, **kwargs)

class detectors.aggregations.anomaly.IFAggregation(n_estimators=100, **kwargs)

Bases: AnomalyBaseAggregation

class detectors.aggregations.anomaly.LOFAggregation(n_neighbors: int = 20, p=2, **kwargs)

Bases: AnomalyBaseAggregation

detectors.aggregations.basics module

detectors.aggregations.basics.avg_topk_aggregation(x: Tensor, k: int, *args, **kwargs) → Tensor

detectors.aggregations.basics.depth_weighted_sum(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.layer_idx(x: Tensor, layer_idx: int = -1, *args, **kwargs) → Tensor

detectors.aggregations.basics.max_aggregation(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.mean_aggregation(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.median_aggregation(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.min_aggregation(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.none_aggregation(x: Tensor, *args, **kwargs) → Tensor

detectors.aggregations.basics.topk_aggregation(x: Tensor, k: int, *args, **kwargs) → Tensor

detectors.aggregations.cosine module

class detectors.aggregations.cosine.CosineAggregation(*args, **kwargs)

Bases: object

fit(stack: Tensor, *args, **kwargs)

detectors.aggregations.innerprod module

class detectors.aggregations.innerprod.InnerProductAggregation(*args, **kwargs)

Bases: object

fit(stack: Tensor, *args, **kwargs)

detectors.aggregations.mahalanobis module

class detectors.aggregations.mahalanobis.MahalanobisAggregation(*args, **kwargs)

Bases: object

fit(stack: Tensor, *args, **kwargs)

detectors.aggregations.mahalanobis.mahalanobis_distance_inv_fast(x: Tensor, y: Tensor, precision: Tensor)

Mahalanobis distance betwee x and y with an accelerated implementation.

Parameters:

• x (Tensor) – first point.

• y (Tensor) – second point.

• precision (Tensor) – inverse of the covariance matrix.

detectors.aggregations.quantile module

class detectors.aggregations.quantile.QuantileAggregation

Bases: object

fit(stack: Tensor, *args, **kwargs)

detectors.aggregations.supervised module

class detectors.aggregations.supervised.WeightRegression(verbose=False, ignore_dim=0, split_size=1000, regressor=<class 'sklearn.linear_model._logistic.LogisticRegressionCV'>, recall_level=0.95, seed=0)

Bases: object

scoring_obj(y_true, y_pred)

split_data(data_in, data_out)

Module contents

class detectors.aggregations.Aggregation(aggregation_method, *args, **kwargs)

Bases: object

Aggregation wrapper class.

fit(stack: Tensor, *args, **kwargs)

detectors.aggregations.create_aggregation(aggregation_name: str, **kwargs) → Aggregation

detectors.aggregations.list_aggregations() → list

detectors.aggregations.register_aggregation(name: str)

Decorator to register a new aggregation method.