[ENH] Matrix profile-based anomaly detectors: left STAMPi

[CodeLionX]

Describe the feature or idea you want to propose

The matrix profile can be computed incrementally, using e.g. STAMPi. If only the distances to all already observed windows are computed, the resulting data structure is called left matrix profile. This allows us to detect reoccurring patterns as anomalies that would otherwise not be detected because of the twin-freak problem. Add such an algorithm to aeon.

References:

• STAMP and STAMPi are described in: Chin-Chia Michael Yeh, Yan Zhu, Liudmila Ulanova, Nurjahan Begum, Yifei Ding, Hoang Anh Dau, Diego Furtado Silva, Abdullah Mueen, and Eamonn Keogh. 2016. Matrix Profile I: All Pairs Similarity Joins for Time Series: A Unifying View That Includes Motifs, Discords and Shapelets. In Proceedings of the International Conference on Data Mining (ICDM), 1317–1322. doi: 10.1109/ICDM.2016.0179.

Describe your proposed solution

• TimeEval implementation: https://github.com/TimeEval/TimeEval-algorithms/blob/main/left_stampi/
• tags: unsupervised, univariate
• compute left matrix profile of the input time series using stumpy as described in their docs
• convert left matrix profile (window-based anomaly score) to point-based anomaly score using aeon.utils.reverse_windowing and window_size=m

Describe alternatives you've considered, if relevant

No response

Additional context

No response