I have got review for my work saying, that my work (adaptive filter variant) should be analyzed in transient and steady-state before claiming it improves performance.

I have done common (in my opinion) analysis:

  1. prediction of linear system
  2. prediction of non-linear system
  3. prediction of real measured non-stationary data

How can I extend my analysis with transient and steady-state experiments? How the experimental data should look like?

Thank you in advance for any hint, note, reference or example. I am pretty sure that I am missing something basic.

  • $\begingroup$ Maybe it would be interesting to analyze how the adaptive filtering algorithm behaves when the system changes. For example, you could let the filter converge, then change the system and don't reset the filter state but let it adapt to the new environment. $\endgroup$
    – applesoup
    Jan 23, 2017 at 15:18
  • $\begingroup$ @applesoup Yes, that sounds really reasonable. If you found any study what is done in this way, please create an answer, I will accept it. $\endgroup$
    – matousc
    Jan 27, 2017 at 19:47
  • $\begingroup$ I couldn't find a reference for this approach yet. I used it for a project and it worked for me, but I will add an answer when I have found a published reference. $\endgroup$
    – applesoup
    Jan 30, 2017 at 12:30


Your Answer

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