dataeval.quality.Outliers.Config¶

class dataeval.quality.Outliers.Config¶

Configuration for Outliers detector.

flags¶

Statistics to compute for image statistics-based outlier detection.

Type:¶

ImageStats, default ImageStats.DIMENSION | ImageStats.PIXEL | ImageStats.VISUAL

outlier_threshold¶

Threshold configuration. When None, uses AdaptiveThreshold(3.5) (Double-MAD with asymmetric bounds). See Outliers for full description.

Type:¶

ThresholdLike | Mapping[str, ThresholdLike] | None, default None

cluster_threshold¶

Threshold configuration for cluster-based detection. When None, defaults to ZScoreThreshold(upper_multiplier=2.5).

Type:¶

ThresholdLike or None, default None

extractor¶

Feature extractor for cluster-based outlier detection.

Type:¶

FeatureExtractor or None, default None

batch_size¶

Batch size for feature extraction during cluster-based detection. If None, uses DataEval default. Must be set by either parameter or global default if extractor is provided.

Type:¶

int or None, default None

cluster_algorithm¶

Clustering algorithm for cluster-based detection.

Type:¶

{“kmeans”, “hdbscan”}, default “hdbscan”

n_clusters¶

Expected number of clusters.

Type:¶

int or None, default None