Published Version 0.2.0

In this release, we:

• Add support for the decomposition of Block Hankel matrices (including fast Versions with the FFT representation)
• Add support for multivariate SST and ESST (MSST & MESST

Published Version 0.1.2

In this release, we:

• Add a new SST submethod ("weighted") that allows to incorporate more than one future singular vector into the score
• Add a new offset_mitigation parameter to SST and ESST that allows translation invariance of the score (score invariance to additive signal offset) at the cost of scoring noise for small window sizes or very noisy signals. The parameter mitigates the effect of the absolut signal values on the score.
• Updated the IKA-SST so it works with the newest numpy version
• Increase the numerical stability of the power method, used to extract the future singular vector, and slightly bump the number of iterations when used in the SST
• Updated the "rsvd" SST to use the rSVD for both future and past singular vectors instead of only for the past singular vectors
• Deleted the parallel ESST processing pipeline as it was brittle within more complex deployments

Published Version 0.1.1

• Added Accelerated SST and IKA-SST publication, and updated documentation accordingly
• minor documentation cosmetics

Published Version 0.1.0

• Added updated naive SST version (score from Idé et al. 2007 but algorithm from Idé et al. 2005)
• Added BOCPD mean change method
• Added some README notes
• Added ESST citation to code and README

Published Version 0.0.9

• Added the baseline methods used for the IAS paper
• Added BOCPD for mean changes

Published Version 0.0.8

This version features the following updates:

• Exclude claspy due to old enforced requirements
• Fix minor mistakes for SST
• include fast Hankel matrix product

Published Version 0.0.7

The new package has been released ap PyPI. It includes bugfixes and mainly two new methods uLSIF and RuLSIF