在这个例子中是否可以使用我创建的所有类共有的超类？答案

【问题标题】：Is it possible to use a super class in this exemple, common to all the classes I created?在这个例子中是否可以使用我创建的所有类共有的超类？
【发布时间】：2020-07-14 16:13:33
【问题描述】：

我有所有这些非常相似的类，除了它们的 covar_module。我想知道是否可以创建一个通用的超类，它具有所有属性和功能，而子类只有 covar_module。我对python很陌生，所以我不知道synthax会怎样。

class RbfGP(ExactGP, GPyTorchModel):
    _num_outputs = 1  # to inform GPyTorchModel API

    def __init__(self, train_X, train_Y):
        # squeeze output dim before passing train_Y to ExactGP
        super().__init__(train_X, train_Y.squeeze(-1), GaussianLikelihood())
        self.mean_module = ConstantMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(
            base_kernel= gpytorch.kernels.RBFKernel(ard_num_dims=train_X.shape[-1]),
        )
        self.to(train_X)  # make sure we're on the right device/dtype

    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return MultivariateNormal(mean_x, covar_x)


class Matern12GP(ExactGP, GPyTorchModel):
    _num_outputs = 1  # to inform GPyTorchModel API

    def __init__(self, train_X, train_Y):
        # squeeze output dim before passing train_Y to ExactGP
        super().__init__(train_X, train_Y.squeeze(-1), GaussianLikelihood())
        self.mean_module = ConstantMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(
            base_kernel=gpytorch.kernels.MaternKernel(nu=0.5, ard_num_dims=train_X.shape[-1]),
        )
        self.to(train_X)  # make sure we're on the right device/dtype

    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return MultivariateNormal(mean_x, covar_x)


class Matern32GP(ExactGP, GPyTorchModel):
    _num_outputs = 1  # to inform GPyTorchModel API

    def __init__(self, train_X, train_Y):
        # squeeze output dim before passing train_Y to ExactGP
        super().__init__(train_X, train_Y.squeeze(-1), GaussianLikelihood())
        self.mean_module = ConstantMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(
            base_kernel=gpytorch.kernels.MaternKernel(nu=1.5, ard_num_dims=train_X.shape[-1]),
        )
        self.to(train_X)  # make sure we're on the right device/dtype

    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return MultivariateNormal(mean_x, covar_x)

【问题讨论】：

标签： python superclass

【解决方案1】：

为什么不这样：

class GP(ExactGP, GPyTorchModel):
    _num_outputs = 1  # to inform GPyTorchModel API

    def __init__(self, train_X, train_Y, base_kernel, **kwargs):
        # squeeze output dim before passing train_Y to ExactGP
        super().__init__(train_X, train_Y.squeeze(-1), GaussianLikelihood())
        self.mean_module = ConstantMean()
        self.covar_module = gpytorch.kernels.ScaleKernel(
            base_kernel=base_kernel(ard_num_dims=train_X.shape[-1], **kwargs),
        )
        self.to(train_X)  # make sure we're on the right device/dtype

    def forward(self, x):
        mean_x = self.mean_module(x)
        covar_x = self.covar_module(x)
        return MultivariateNormal(mean_x, covar_x)

然后

rbf_gp = GP(train_x, train_y, base_kernel=gpytorch.kernels.RBFKernel)

matern_12_gp = GP(train_x, train_y, base_kernel=gpytorch.kernels.MaternKernel, nu=0.5)

【讨论】：