This repository has been archived by the owner on Feb 28, 2024. It is now read-only.
Gaussian process: Formula for covariance. #1112
Unanswered
dk-teknologisk-cbb
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
I have noticed something that could be an inconsistency between this repository and the book Gaussian Processes for Machine Learning (GPML) by Rasmussen and Williams.
On p. 19 in GPML, Algorithm 2.1 line 6 calculates the variance in the following way:
In gpr.py l. 322 looks like this:
y_cov = self.kernel_(X) - K_trans.dot(v) # Line 6
Notice that v transposed has been replaced by K transposed in the code.
In scikit learn, _gpr.py (l. 409) has a different formula that seems consistent with the book:
y_cov = self.kernel_(X) - V.T @ V
Do any of you know if this implementation detail is correct in scikit optimize?
If it is correct, what is the explanation for the difference between the book and this repository?
Beta Was this translation helpful? Give feedback.
All reactions