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
Use this guide for a coarse-to-fine parameter search, runtime estimation, decomposition complexity, and fast Hankel decisions for SST, ESST, MSST, and MESST.
Use this focused guide when large Hankel matrices dominate the runtime of SST-style methods.
Visualization
Plot a signal and its change score together, using background shading to align high-score regions with the original data.
Experimental Streaming
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:
ULSIFandRuLSIF. - Matrix-profile segmentation:
FLUSSandFLOSS. - Probabilistic methods:
BOCPD.
Each guide can compare related algorithms, explain their shared parameters, and link to the individual classes in the API reference.