stumpy.match bug with parameters: Sliding mean and Sliding standard deviation

[brunopinos31]

I have a bug when i try to use the stumpy.match function with parameters: Sliding mean and Sliding standard deviation.

[seanlaw]

@brunopinos31 Thank you for reporting this and welcome to the STUMPY community. I will take a look

[brunopinos31]

I mean the length of the motif.

For stumpy.motif:
input: pan matrix profile computed with stimp
output: motifs of different length

[seanlaw]

So, unless you are computing the pan matrix profile with percentage=1.0 (e.g., stumpy.stimp(T, min_m=3, max_m=10, percentage=1.0)), I don't think it's a good idea to be using the pan matrix profile to find motifs as any value of percentage < 1.0 is only an approximation of the exact matrix profile. As a consequence, any motifs obtained from an approximate pan matrix profile would also be approximate. However, if you are are computing the pan matrix profile with percentage=1.0 then you might as well just do:

mps = []
for m in range(3, 11):
    mps.append(stumpy.stump(T, m))

and then apply stumpy.motifs to each matrix profile.

[brunopinos31]

Ok thx Seanlaw.
I'm looking for an implementation of VALMOD in python.
I don't know if it exists^^

[seanlaw]

You're correct, we do not have a VALMOD implementation. There is also this other related paper on MAD. I haven't read them yet but it may be relevent

[brunopinos31]

I may try to do something for valmod I'll let you know. Thank you for all !