tawheeler
commented
2 years ago Good question. From Gaussian Processes for Machine Learning:
The noise is only added to the noisy observations, not the true function values.
Closed alextzik closed 2 years ago
tawheeler
commented
2 years ago Good question. From Gaussian Processes for Machine Learning:
The noise is only added to the noisy observations, not the true function values.
Why is noise only included in the covariance of the measurements we will be conditioning on in eq. (15.28) (y), given that we will be evaluating the same function to get y_hat? Shouldn't we include the term v*I in the upper-left matrix of the covariance also?