Using sparse GP regression I’ve got a target that has got known noise covariance that is different for each target. Is it possible to to use a deterministic noise + a white noise covariance with the marginalsparse + marginal_likehood methods?
Right now it is not, though it’s possible to do. Easily for DTC, and I’ve seen a derivation of it for VFE somewhere, not sure about FITC, would need to check.