My task is to classify into two classes the time series like these shown in the figure.

Time series divided by two classes (left and right).

The figure shows one class on the left sub-figure and second one on the right. The series are shown in pairs for more clarity, but each series on the left (and right) belongs to one of the respective classes. The scale of the right panel is reduced to show all series, but these series are of the same amplitude as the left ones.

Is it possible to apply RNN (or other methods) to classify the series of this kind into two classes?

I have never used neural networks, but I am just looking for an adequate method for this problem.

  • $\begingroup$ Yes this is definitely possible, considering you have enough curve with labels. You should maybe go through a dummy example of RNN or LSTM (in Keras or Pytorch) to familiarize yourself with this. $\endgroup$
    – Romain
    Mar 26 at 13:02
  • $\begingroup$ @Romain Thanks a lot. I will familiarize with these method. Could you estimate a number of cases that I should have for teaching? $\endgroup$
    – Marek
    Mar 26 at 13:10
  • $\begingroup$ Impossible to say as this is problem dependent, but from my personal experience I would say in the 100s or 1000s. Maybe someone else has a better idea than me on that part. $\endgroup$
    – Romain
    Mar 26 at 13:14
  • $\begingroup$ This quantity seems to be available in my case (but not readily :). Thank you very much for you advice. $\endgroup$
    – Marek
    Mar 26 at 13:28

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.