Bayesian Gaussian process regression.
Calls tgp::bgp() from package tgp.
Predictions return the posterior predictive mean and the square root of the
predictive kriging variance as standard error. Several parameters that
appear in both tgp::bgp() and predict.tgp are prefixed with predict_
to distinguish the predict-time variants (e.g. predict_BTE,
predict_R).
The learner always trains without collecting training-location predictive summaries and always predicts with pointwise kriging variances only. The corresponding upstream options are therefore not exposed because their effects are discarded by the mlr3 wrapper.
Creates a new instance of this learner.