English

Conditional Information Inequalities and Combinatorial Applications

Information Theory 2017-09-14 v4 math.IT

Abstract

We show that the inequality H(AB,X)+H(AB,Y)H(AB)H(A \mid B,X) + H(A \mid B,Y) \le H(A\mid B) for jointly distributed random variables A,B,X,YA,B,X,Y, which does not hold in general case, holds under some natural condition on the support of the probability distribution of A,B,X,YA,B,X,Y. This result generalizes a version of the conditional Ingleton inequality: if for some distribution I(X:YA)=H(AX,Y)=0I(X: Y \mid A) = H(A\mid X,Y)=0, then I(A:B)I(A:BX)+I(A:BY)+I(X:Y)I(A : B) \le I(A : B \mid X) + I(A: B \mid Y) + I(X : Y). We present two applications of our result. The first one is the following easy-to-formulate combinatorial theorem: assume that the edges of a bipartite graph are partitioned into KK matchings such that for each pair (left vertex xx, right vertex yy) there is at most one matching in the partition involving both xx and yy; assume further that the degree of each left vertex is at least LL and the degree of each right vertex is at least RR. Then KLRK\ge LR. The second application is a new method to prove lower bounds for biclique coverings of bipartite graphs.

Keywords

Cite

@article{arxiv.1501.04867,
  title  = {Conditional Information Inequalities and Combinatorial Applications},
  author = {Tarik Kaced and Andrei Romashchenko and Nikolay Vereshchagin},
  journal= {arXiv preprint arXiv:1501.04867},
  year   = {2017}
}
R2 v1 2026-06-22T08:07:16.836Z