Skip to content

User Guides

The algorithm overview lists what is available. The guides focus on choosing and configuring an algorithm for a particular kind of change.

Subspace Methods

Tuning subspace methods

Use this guide for a coarse-to-fine parameter search, runtime estimation, decomposition complexity, and fast Hankel decisions for SST, ESST, MSST, and MESST.

Fast SST and IKA-SST

Use this focused guide when large Hankel matrices dominate the runtime of SST-style methods.

Visualization

Visualizing change scores

Plot a signal and its change score together, using background shading to align high-score regions with the original data.

Experimental Streaming

Buffered stream processing

Experiment with repeatedly applying the existing batch transform() interface to a fixed-size rolling buffer and emitting the newly available delayed score.

Adding Further Guides

Future guides should be grouped by method family rather than by individual class. Suitable groups for the current package are:

  • Density-ratio methods: ULSIF and RuLSIF.
  • Matrix-profile segmentation: FLUSS and FLOSS.
  • Probabilistic methods: BOCPD.

Each guide can compare related algorithms, explain their shared parameters, and link to the individual classes in the API reference.