|
As posted in #349 , Is there a way to just retrieve the distance matrix, so that one can design a way to retrieve the top-K matches for EVERY sub sequence within their time series. ?? |
Answered by
seanlaw
Jun 12, 2021
Replies: 1 comment 3 replies
|
@iamdavie No, unfortunately, it is not possible to return the distance matrix with STUMPY as this is beyond goals/scope of computing matrix profiles according to the original authors of the matrix profile work. For a time series with length |
3 replies
Answer selected by
iamdavie
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
@iamdavie No, unfortunately, it is not possible to return the distance matrix with STUMPY as this is beyond goals/scope of computing matrix profiles according to the original authors of the matrix profile work. For a time series with length
n, storing the distance matrix requiresO(n*n)space. So, even a short time series wheren = 100,000this would require roughly 80 GB of memory. The point of the matrix profile data structure is to beO(n)in space complexity.