I have a sparse dataset indexed by nanoseconds. Storing the dataset in a discrete fashion would take too much memory. I'd like to take a wavelet transform and I'd like it to be relatively fast. The dataset has about 100,000 dirac deltas in it.

Is this possible?

Thanks! James

  • $\begingroup$ Welcome. I do not understand the link between your question and the tags you have chosen. Then, the purpose of using a wavelet transform on already sparse data is not fully clear to me. Could you please elaborate on the objective? $\endgroup$ Nov 23 '20 at 16:22
  • 1
    $\begingroup$ Sorry, the tags were from a draft of something else. I've fixed it! $\endgroup$ Nov 23 '20 at 16:34
  • $\begingroup$ What is the typical of each stride of data? Shall we assume that the deltas are of constant of variable amplitude? Are the other samples zeros, or is there noise? $\endgroup$ Nov 23 '20 at 18:52
  • $\begingroup$ The data has a fractal quality to it. There can be short periods where it changes a large amount in amplitude and the frequency of points varies with those changes. The data spans 6 hours and is marked with nanoseconds (or microseconds works too). Outside of these 100k events, everything else is 0 (in delta space). The 100k events represent a change in the underlying value. I have 1 million different time series all like this and I want to take a wavelet transform of all of them in an efficient manner (on a distributed cluster). $\endgroup$ Nov 23 '20 at 19:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.