We introduce a prior for the parameters of univariate continuous distributions, based on the Wasserstein information matrix, which is invariant under reparameterisations. We briefly discuss the links between the proposed prior with information geometry. We present several examples where we can either obtain this prior in closed-form, or propose a numerically tractable approximation for cases where the prior is not available in closed-form. Since this prior is improper in some cases, we present sufficient conditions for the propriety of the posterior distribution for general classes of models.
翻译:我们根据Wasserstein信息矩阵(该矩阵在重新校正中是不变的),对单体连续分布参数引入了前置参数。我们简要讨论了先前建议与信息几何之间的联系。我们举了几个例子,我们既可以在封闭式之前获得,也可以在封闭式情况下提出数字可移动近似值。由于前一种情况在某些情况下是不恰当的,我们为一般类型的模型的后置分布提供了充分的条件。