Controlling subsequence comparisons

[gerold-csendes-epam]

Hello stumpy folks,
let's say I have a task for finding motifs where m=168 because we assume weekly (hour-level) pattern. For a downstream task, it may be useful to find motifs that match "week-by-week". So, instead of comparing all subsequences, we just compare specific ones e.g. Monday-Sunday to Monday-Sunday.
Is this possible in stumpy? I know this is a really specific task and after reading the docs I don't think it can be done. Still, I would be happy if someone could verify this.
Thanks a lot! :)

[seanlaw]

Hello stumpy folks,
let's say I have a task for finding motifs where m=168 because we assume weekly (hour-level) pattern. For a downstream task, it may be useful to find motifs that match "week-by-week". So, instead of comparing all subsequences, we just compare specific ones e.g. Monday-Sunday to Monday-Sunday.
Is this possible in stumpy? I know this is a really specific task and after reading the docs I don't think it can be done. Still, I would be happy if someone could verify this.
Thanks a lot! :)

@gerold-csendes-epam Before I can provide a response, I'd like to understand your question a little bit better using a more concrete example. Let's say we have 3 weeks worth of hourly data (i.e., 3 x 168 = 504 data points in total) stored in the data array. So, for the first week (i.e., data[ : 168]), you only want to compare (i.e., compute the distance between ) data[ : 168] to data[168:336] and then to data[336:]. Then, you take the min distance. Next, you repeat the same thing for data[168:336] and you compare this subsequence with data[ : 168] and data[336:]. Finally, you repeat the same thing for data[336:] and you compare this subsequence with data[ : 168] and data[168:336]. So, in total, there are only 6 distances being computed and no more. Is that correct? Otherwise, please clarify

[gerold-csendes-epam]

That is an accurate description, yes! Thanks @seanlaw

[seanlaw]

Is this possible in stumpy? I know this is a really specific task and after reading the docs I don't think it can be done. Still, I would be happy if someone could verify this.

You are correct. This cannot be done with STUMPY and you are better off writing a nested for-loop to handle this since there is probably little performance gain to be had.

[gerold-csendes-epam]

great, that is good to know. Thanks a lot!

[seanlaw]

@gerold-csendes-epam Just an FYI that the new version was released today!