1
$\begingroup$

I would like to ask for advice on choosing a suitable modelling method for the following problem:

I am modeling the performance of a device for curve estimation. I have collected a data set consisting of $3^{rd}$ degree polynomial curves (red), where the true $3^{rd}$ degree polynomial curve is in bold green. I need to model the true curve with uncertainty at each point, like shown on the second part of the figure*.
Note, that the resulting model has increasing variance as the curve estimation is less accurate the further away from the device the data is acquired.

*Please excuse the comically crude drawing Figure of data and desired model

I have some machine learning background, so my best (and only) guess would be to use a Gaussian Process regression, which seems suitable since it results in a model with included uncertainty. However, the data I have seems more like samples from the posterior of the GP, and not data over which I could do regression.

So, I am hoping to hear your opinions on which modelling methods are most suitable and why.

$\endgroup$
1
  • $\begingroup$ Why not just take the mean and covariance of the polynomial coefficients? $\endgroup$ Feb 22, 2018 at 8:48

0

Your Answer

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

Browse other questions tagged or ask your own question.