Gaussian Processes - combining constant features with covariance

I may be mistaken, but shouldn’t the Constant cov function not depend on input_dim or active_dims? It should just add a constant across the whole matrix?

K = k(x, x') + c 

That’s why input_dim and active_dims aren’t able to be set. Do you get the covariance structure you’re after this way? If the covariance function is scaled by a constant, you can just do

K = c*k(x, x')

without using the Constant covariance object.