rapid_models.gp_models.templates

Module Contents

Classes

ExactGPModel

Model for standard GP regression

class rapid_models.gp_models.templates.ExactGPModel(train_x: torch.Tensor, train_y: torch.Tensor, mean_module: gpytorch.means.Mean, covar_module: gpytorch.kernels.Kernel, likelihood: gpytorch.likelihoods.Likelihood, path: str = '', name: str = '')

Bases: gpytorch.models.ExactGP

Model for standard GP regression

forward(x: torch.Tensor) → gpytorch.distributions.MultivariateNormal

eval_mode()

Set model in evaluation mode

train_mode()

Set in training mode

predict(x: torch.Tensor, latent: bool = True, CG_tol: float = 0.1, full_cov: bool = False) → Tuple[torch.Tensor, torch.Tensor]

Return mean and covariance at x

Input: x - tensor of size dim * N containing N inputs latent - latent = True -> using latent GP

latent = False -> using observed GP (incl. likelihood)

CG_tol - Conjugate Gradient tolerance for evaluation full_cov - full_cov = False -> Return only diagonal (variances)

Output: mean and covariance

print_parameters()

Print actual (not raw) parameters

save()

Save GP model parameters to self.path

load()

Load GP model parameters from self.path