Austin Ouyang

https://github.com/target/matrixprofile-ts/blob/bcba7dc741d254435a72a60d6e014bf563de7d5a/matrixprofile/utils.py#L154 I believe is the culprit here. Using the precomputed cache to populate the 0th index of the next row's distance profile makes the assumption that we're doing a self-join....

@ortegatron You can also take a look at this paper for multivariate mp. https://www.cs.ucr.edu/~eamonn/Motif_Discovery_ICDM.pdf.

@oentaryorj Something that's worked for me was to add some very small noise to the signal. This has the added benefit that the flat regions can be detected later on...

Agreed. Found this from stack overflow as well https://github.com/scikit-learn/scikit-learn/blob/7389dbac82d362f296dc2746f10e43ffa1615660/sklearn/preprocessing/data.py#L70 treating the standard deviation as 1 keeping the existing subsequence values and not normalizing. Since we're trying to compare to subsequences...

> I've taken a brief look through this. There is more API here that I think is necessary. Also there are error returns that do not conform to the approach...

> The API is too complex here. I'd estimate that we could make this somewhere between 1/3 and 2/3 of the code that's here by removing the extraneous code. Sounds...

> Hi, thank you guys for the work you've put in. What's the status of this PR, because I'd really like to see this feature in the main branch. It's...