Hi David, thanks for the nice package! It is really helpful!
I plan to apply your package to integer parameters, which might require playing some tricks on the kernel (e.g., adding a white noise).
Here I have a question relating to the code at approx.py#L712-L716
It creates a new GP object, however, it does not hold all the original properties in self.gp, e.g., fit_white_noise can be True in a user-supplied gp.
While we can check the status in gp.models["white_noise"].unfrozen_mask in principle, it seems the creation of a new GP object here is not necessary. Because the following line will update all the things we need (as far as I see)
self.gp.compute(self.theta)
If the above understanding is correct, it sounds reasonable to remove lines L712-L716.
Am I right? Or I have missed something? Thanks!
Hi David, thanks for the nice package! It is really helpful!
I plan to apply your package to integer parameters, which might require playing some tricks on the kernel (e.g., adding a white noise). Here I have a question relating to the code at approx.py#L712-L716
It creates a new GP object, however, it does not hold all the original properties in
self.gp
, e.g., fit_white_noise can be True in a user-supplied gp.While we can check the status in
gp.models["white_noise"].unfrozen_mask
in principle, it seems the creation of a new GP object here is not necessary. Because the following line will update all the things we need (as far as I see)If the above understanding is correct, it sounds reasonable to remove lines L712-L716. Am I right? Or I have missed something? Thanks!