Refined Pinsker's and reverse Pinsker's inequalities for probability distributions of different dimensions

Michele Caprio
DOI: https://doi.org/10.48550/arXiv.2203.00500
2022-11-02
Abstract:We provide optimal lower and upper bounds for the augmented Kullback-Leibler divergence in terms of the augmented total variation distance between two probability measures defined on two Euclidean spaces having different dimensions. We call them refined Pinsker's and reverse Pinsker's inequalities, respectively.
Statistics Theory,Probability
What problem does this paper attempt to address?