b61a525baee25d2ea31ab088540dd4d2ffed2251,test/examples/spectral_mixture_gp_regression_test.py,SpectralMixtureGPModel,__init__,#SpectralMixtureGPModel#,22

Before Change

class SpectralMixtureGPModel(gpytorch.GPModel):
    def __init__(self):
        super(SpectralMixtureGPModel, self).__init__(GaussianLikelihood())
        self.mean_module = ConstantMean()
        self.covar_module = SpectralMixtureKernel()
        self.register_parameter("log_noise", nn.Parameter(torch.Tensor([-2])), bounds=(-5, 5))
        self.register_parameter("log_mixture_weights", nn.Parameter(torch.zeros(3)), bounds=(-5, 5))
        self.register_parameter("log_mixture_means", nn.Parameter(torch.zeros(3)), bounds=(-5, 5))
        self.register_parameter("log_mixture_scales", nn.Parameter(torch.zeros(3)), bounds=(-5, 5))

After Change

    def __init__(self):
        likelihood = GaussianLikelihood(log_noise_bounds=(-5, 5))
        super(SpectralMixtureGPModel, self).__init__(likelihood)
        self.mean_module = ConstantMean(constant_bounds=(-1, 1))
        self.covar_module = SpectralMixtureKernel(
            n_mixtures=3,
            log_mixture_weight_bounds=(-5, 5),

Instances