Source code for spateo.tdr.interpolations.interpolation_gaussianprocess.gp_models

import gpytorch
from gpytorch.models import ApproximateGP, ExactGP
from gpytorch.variational import CholeskyVariationalDistribution, VariationalStrategy


[docs]class Approx_GPModel(ApproximateGP):
    def __init__(self, inducing_points):
        variational_distribution = CholeskyVariationalDistribution(inducing_points.size(0))
        variational_strategy = VariationalStrategy(
            self, inducing_points, variational_distribution, learn_inducing_locations=True
        )
        super(Approx_GPModel, self).__init__(variational_strategy)
        self.mean_module = gpytorch.means.ConstantMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(gpytorch.kernels.RBFKernel())

[docs]    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return gpytorch.distributions.MultivariateNormal(mean_x, covar_x)


[docs]class Exact_GPModel(ExactGP):
    def __init__(self, train_x, train_y, likelihood):
        super().__init__(train_x, train_y, likelihood)
        self.mean_module = gpytorch.means.ZeroMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(gpytorch.kernels.RBFKernel())

[docs]    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return gpytorch.distributions.MultivariateNormal(mean_x, covar_x)