Skip to content

Models API

Model documentation is split by event structure and method family.

  • Core utility models

    Static/adaptive thresholds and null/random baselines.

  • Isolation forests

    Incremental, window-replaced, half-space, Mondrian, random-cut, histogram, and xStream variants.

  • Distance and neighborhood

    KNN, LOF, observer-distance, and point-scoring cell adaptations.

  • Sketch detectors

    MStream, streaming LODA, and streaming RS-Hash adaptations.

  • Graph streams

    AnoEdge-L, ISCONNA, MIDAS, and signed graph sketches.

  • Time series

    Pure-online X-Lag Amnesic DAMP.

  • Statistical models

    Candidate-induced changes in moving univariate/bivariate statistics and squared Mahalanobis distance.

  • SVM models

    Experimental budgeted and graph-gated one-class heuristics.

  • Reconstruction models

    NumPy online autoencoder ensemble and optional user-supplied PyTorch autoencoder.

All model classes expose learn_one and score_one, but their event schemas, readiness conditions, state bounds, and score scales differ. Start with the model guide when selecting a class.