Surrogate Models

smt_optim.surrogate_models

Surrogate Models#

class SmtAutoModel(**kwargs)[source]#

Bases: Surrogate

predict_values(x_pred: ndarray) → ndarray[source]#
predict_variances(x_pred: ndarray) → ndarray[source]#
train(xt: list[ndarray], yt: list[ndarray], **kwargs) → None[source]#

Train the GP on the training data.

Parameters:

  • xt (list[np.ndarray]) – training data variables

  • yt (list[np.ndarray]) – training data values

class SmtGPX(**kwargs)[source]#

Bases: Surrogate

predict_values(x_pred: ndarray) → ndarray[source]#
predict_variances(x_pred: ndarray) → ndarray[source]#
train(xt: list[ndarray], yt: list[ndarray], **kwargs) → None[source]#

Train the GP on the training data.

Parameters:

  • xt (list[np.ndarray]) – training data variables

  • yt (list[np.ndarray]) – training data values

class SmtMFCK(**kwargs)[source]#

Bases: Surrogate

predict_level_covariances(x: ndarray, lvli: int, lvlj: int = None)[source]#

Compute the covariance between two fidelity levels at location x.

Parameters:

  • x (np.ndarray) – Array with the inputs for make the prediction.

  • lvli (int) – First fidelity level.

  • lvlj (int) – Second fidelity level. If not specified, will be set to the highest fidelity level.

Returns:

covariances – Returns the posterior covariance.

Return type:

np.array

predict_values(x_pred: ndarray) → ndarray[source]#
predict_variances(x_pred: ndarray) → ndarray[source]#
train(xt: list, yt: list, **kwargs)[source]#
class Surrogate[source]#

Bases: ABC

Abstract class for surrogate models.

abstractmethod predict_values(x_pred: ndarray) → ndarray[source]#
abstractmethod predict_variances(x_pred: ndarray) → ndarray[source]#
abstractmethod train(xt: list[ndarray], yt: list[ndarray], **kwargs) → None[source]#