Closed RunzeZhang123 closed 3 months ago
Hey Runze,
My reasoning was that the parameters estimated during the evolutionary process deal with a trade-off between not taking too long and getting a good fit. The optimisation in the parameter optimisation
function does a more thorough optimisation, which takes more time. I expected it to slightly further improve the fit of the final circuit, though this may not always be the case.
I removed the finetuning step, such that the parameters obtained through GEP are returned
Hey Maxime,
We noticed that when outputting the elite from the evolution process, you used parameteroptimisation to refit the best candidate. Could you please explain why that is necessary? In my opinion, the refit process may alter the circuit fitness to make it not the most competitive one anymore because of the multi-linear fit algorithms' stochastic behavior. Would it be better to just use the parameters estimated during the evolutionary process?
EquivalentCircuit(best_circuit,parameteroptimisation(best_circuit,measurements,frequencies))
Thanks!