MIM4DD: Mutual Information Maximization for Dataset Distillation

A.1 In-variance of Mutual Information Theorem 1 (In-variance of Mutual Information): Mutual information is invariant under reparametrization of the marginal variables.