Detectors

tsseg.algorithms exposes 30+ segmentation algorithms behind a common fit / predict / fit_predict API derived from aeon.segmentation.base.BaseSegmenter.

Each detector declares its capabilities through tags:

capability:univariate / capability:multivariate
returns_dense — True for state labels, False for sparse change points
semi_supervised — True when the algorithm accepts labels at fit time

Use the table of contents below to jump to a specific detector. The detectors are grouped into two tasks: change-point detection returns sparse breakpoint indices, state detection returns dense state labels.

Change-point detection

State detection

Baseline

Random (baseline)

Conventions

Note

The fit signature is fit(X, y=None). For unsupervised algorithms y must be None; passing it is reserved for semi-supervised or supervised detectors that explicitly support it.

Parameters that used to be inferred from y (such as n_segments or n_states) must now be passed explicitly to the constructor.

See Contributions for the full detector-side contract (tags, output shapes and the estimator-check test suite that every algorithm in __all__ must pass).