As stated in the title, is there a way to adapt PSO to an online scenario where new data samples arrive continuously?

In more detail: suppose that I have a classifier with several parameters for which the optimal values are to be chosen automatically, instead of being predefined. I want to use PSO to select the parameters. I know this is doable in a static scenario, where the data set is fixed. However, if new data samples arrive over time (and in large amounts), is there a way to make PSO work on such dynamic data streams?

Also, I am open to other ways to implement self-adaptive parameters. PSO is a possible choice but if it's not possible I'd love to hear your suggestions about other approaches.

  • $\begingroup$ Hello. By "parameters", do you mean "hyper-parameters" (e.g. learning rate in stochastic gradient descent)? PSO is sometimes used for hyper-parameter optimization and you tagged the question with that tag, so that's why I am asking this question. Which parameters are you optimizing specifically and would you need to optimize in an online setting (and why)? $\endgroup$
    – nbro
    Nov 29 '21 at 15:59
  • $\begingroup$ a latest survey may hint you the existing approaches to using evolutionary algorithm (e.g., pso) to solve dynamic optimization problem : doi.org/10.1109/TEVC.2021.3060014 $\endgroup$
    – Sanyou
    Nov 29 '21 at 16:12
  • 1
    $\begingroup$ Thank you both for your comments! Yes Sanyou I'll take a look at it thanks for the reference! And nbro specifically I want to perform clustering on data streams and the algorithm I want to use has several hyper-parameters that are normally predefined based on some pre-analysis of a data stream. I want to set the hyper-parameters automatically instead. So that's why I'm trying to see if PSO is applicable in this case. $\endgroup$
    – Elise Le
    Nov 29 '21 at 20:05

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.