MODULE REFERENCE

generic

anomx.detection.generic

Explicit adapters for user supplied scoring and detection algorithms.

GenericDetector

GenericDetector(model: Any = None, *, detect_fn: Callable | None = None, fit_fn: Callable | None = None, anomaly_label: int = 1)

Bases: Detector

Wrap a detector callable or estimator and normalize binary labels.

detect_fn(model, values) overrides detect/predict. Output must have one row per input row. By default labels are bool or {0, 1}; set anomaly_label=-1 for sklearn's {-1, 1} convention. Unknown labels raise instead of silently changing their meaning. DataFrame results add an is_anomaly column, while TimeSeries results are numeric binary series.

Signature parameters
Parameter	Type	Default
`model`	`Any`	`None`
`detect_fn`keyword only	`Callable \| None`	`None`
`fit_fn`keyword only	`Callable \| None`	`None`
`anomaly_label`keyword only	`int`	`1`

Members

component_keyattribute

# GenericDetector.component_key Source ↗

component_key = 'generic_detector'

component_discoverableattribute

# GenericDetector.component_discoverable Source ↗

component_discoverable = False

fitmethod

# GenericDetector.fit Source ↗

fit(data: Any) -> GenericDetector

Fit the custom detector explicitly on training data.

Signature parameters
Parameter	Type	Default
`data`	`Any`	`required`

Returns GenericDetector

detectmethod

# GenericDetector.detect Source ↗

detect(data: Any) -> Any

Return binary labels with exactly the input observation alignment.

Signature parameters
Parameter	Type	Default
`data`	`Any`	`required`

Returns Any

Inherited members

classView source ↗

GenericScorer

GenericScorer(model: Any = None, *, score_fn: Callable | None = None, fit_fn: Callable | None = None, higher_is_anomaly: bool = True)

Bases: Scorer

Wrap a scoring callable or estimator with explicit anomaly direction.

The callable receives a finite numeric array with observations on axis 0 and returns one scalar or vector per observation. Alternatively provide an estimator with score_samples, decision_function, or score. score_fn(model, values) overrides method selection. Set higher_is_anomaly=False for sklearn density/inlier scores; the adapter negates those outputs. For DataFrames, a scalar output is added as score; TimeSeries outputs preserve time coordinates. Fitting, when available, is always explicit.