custom_GP

Custom implementations of Gaussian Process models using BoTorch API

Classes

DKLGP(train_X, train_Y)	GP surrogate with a FF NN feature extractor
FlatGP(train_X, train_Y[, nu])	GP Surrogate with non-informative or no prior distributions
PriorGP(train_X, train_Y)	Class which builds a GP with custom prior distributions; by default set to the values of BoTorch SingleTaskGP

class obsidian.surrogates.custom_GP.DKLGP(train_X, train_Y)

Bases: ExactGP, GPyTorchModel

GP surrogate with a FF NN feature extractor

forward(x)

Evaluate the forward pass of the model on inputs X

class obsidian.surrogates.custom_GP.FlatGP(train_X, train_Y, nu=2.5)

Bases: ExactGP, GPyTorchModel

GP Surrogate with non-informative or no prior distributions

forward(x)

Evaluate the forward pass of the model on inputs X

class obsidian.surrogates.custom_GP.PriorGP(train_X, train_Y)

Bases: ExactGP, GPyTorchModel

Class which builds a GP with custom prior distributions; by default set to the values of BoTorch SingleTaskGP

forward(x)

Evaluate the forward pass of the model on inputs X