Discrete Entropy Power Inequalities via Sperner Theory

12/04/2017

∙

by   Mokshay Madiman, et al.

∙

0

∙

share

A ranked poset is called strongly Sperner if the size of k-family cannot exceed the sum of k-largest Whitney numbers. In a sense of function ordering, a function f is (weakly) majorized by g if the the sum of k-largest values in f cannot exceed the sum of k-largest values in g. Two definitions arise from different contexts, but each share a strong similarity with each other. Furthermore, the product of two weighted posets assumes a structural similarity with a convolution of two functions. Elements in the product of weighted posets with ranks capture underlying structures of the building blocks in the convolution. Combining all together, we are able to derive several types of entropy inequalities including discrete entropy power inequalities, and discuss more applications.