FitCriterion

class baybe.surrogates.gaussian_process.components.fit_criterion.FitCriterion[source]

Bases: Enum

Available fitting criteria for GP hyperparameter optimization.

Public methods

to_gpytorch(likelihood, model)

Create the corresponding GPyTorch MLL object.

Public attributes and properties

MARGINAL_LOG_LIKELIHOOD

Exact marginal log-likelihood.

LEAVE_ONE_OUT_PSEUDOLIKELIHOOD

Leave-one-out cross-validation pseudo-likelihood.

to_gpytorch(likelihood: GPyTorchLikelihood, model: GPyTorchModel)[source]

Create the corresponding GPyTorch MLL object.

Return type:

MarginalLogLikelihood

LEAVE_ONE_OUT_PSEUDOLIKELIHOOD = 'LEAVE_ONE_OUT_PSEUDOLIKELIHOOD'

Leave-one-out cross-validation pseudo-likelihood.

MARGINAL_LOG_LIKELIHOOD = 'MARGINAL_LOG_LIKELIHOOD'

Exact marginal log-likelihood.