Regression using RBF kernel

Both - but you can think of them as the uncertainty from different source. Notice that when you are working with Gaussian likelihood, the noise is the variance added to the diagonal of the cov matrix.

1 Like