0
$\begingroup$

I've been reading the paper https://arxiv.org/pdf/1906.03821.pdf for spectral residual outlier detection, but I don't quite understand it. Specifically, in the implementation there are three variables that determine how to algorithm will find outliers:

# less than period
amp_window_size

# (maybe) as same as period
series_window_size

# a number enough larger than period
score_window_size

Also, the only documentation that I could find on the algorithm was this.

Can anybody familiar with the paper, or with time series, explain how the algorithm works, and how these 3 variables determine how outliers are detected?

$\endgroup$

1 Answer 1

1
$\begingroup$

Can I make a slightly more general observation?

Look at the datasets they test on, and read [a].

There is no read evidence that this idea works, so why bother to implement it?

[a] Current Time Series Anomaly Detection Benchmarks are Flawed and are Creating the Illusion of Progress https://arxiv.org/abs/2009.13807

$\endgroup$
1
  • $\begingroup$ Your answer could be improved with additional supporting information. Please edit to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers in the help center. $\endgroup$
    – Community Bot
    Commented Jan 18, 2022 at 20:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.