Projections, entropy and sumsets (Q452821): Difference between revisions

Given a sequence of \(n\) random variables \(X=(X_1,\dots,X_n)\), let \(X_A=(X_i)_{i\in A}\) for any \(A\subset\{1,\dots,n\}\). In the first part of the paper under review, the authors obtain an entropy inequality which extends previous results by Shearer and \textit{M. Madiman} and \textit{P. Tetali} [``Sandwich bounds for joint entropy'', in: Proceedings of the IEEE International Symposium on Information Theory, ISIT 2007. France: IEEE (2007; \url{doi:10.1109/ISIT.2007.4557276}); ``Information inequalities for joint distributions, with interpretations and applications '', IEEE Trans. Inf. Theory 56, No. 6, 2699-2713 (2010; \url{doi:10.1109/TIT.2010.2046253})], namely, that if \(\mathcal{A},\mathcal{B}\) are finite families of subsets of \(\{1,\dots,n\}\) with \(\mathcal{A}>\mathcal{B}\) (for a suitably defined partial order), then \(\sum_{A\in\mathcal{A}}H(X_A)\geq\sum_{B\in\mathcal{B}}H(X_B)\); here \(H(X)\) denotes the entropy of the random variable \(X\). These entropy results are related to famous projection inequalities like the well-known Loomis-Whitney inequality or the Box Theorem. The second part of the paper is devoted to obtain extensions of inequalities concerning sums of sets of integers, which will be consequences of some projection inequalities.