František Matúš

Infinitely Many Information Inequalities

When finite, Shannon entropies of all sub vectors of a random vector are considered for the coordinates of an entropic point in Euclidean space. A linear combination of the coordinates gives rise to an unconstrained information inequality if it is nonnegative for all entropic points. With at least four variables no finite set of linear combinations generates all such inequalities. This is proved by constructing explicitly an infinite sequence of new linear information inequalities and a curve in a special geometric position to the halfspaces defined by the inequalities. The inequalities are constructed recurrently by adhesive pasting of restrictions of polymatroids and the curve ranges in the closure of a set of the entropic points.

Information projections revisited

The goal of this paper is to complete results available about I-projections, reverse I-projections, and their generalized versions, with focus on linear and exponential families. Pythagorean-like identities and inequalities are revisited and generalized, and generalized maximum-likelihood (ML) estimates for exponential families are introduced. The main tool is a new concept of extension of exponential families, based on our earlier results on convex cores of measures.

Two Constructions on Limits of Entropy Functions

The correspondence between the subvectors of a random vector and their Shannon entropies gives rise to an entropy function. Limits of the entropy functions are closed to convolutions with modular polymatroids, and when integer-valued also to free expansions. The problem of description of the limits of entropy functions is reduced to those limits that correspond to matroids. Related results on entropy functions are reviewed with regard to polymatroid and matroid theories, and perfect and ideal secret sharing

Adhesivity of polymatroids

Two polymatroids are adhesive if a polymatroid extends both in such a way that two ground sets become a modular pair. Motivated by entropy functions, the class of polymatroids with adhesive restrictions and a class of selfadhesive polymatroids are introduced and studied. Adhesivity is described by polyhedral cones of rank functions and defining inequalities of the cones are identified, among them known and new non-Shannon type information inequalities for entropy functions. The selfadhesive polymatroids on a four-element set are characterized by Zhang-Yeung inequalities.

Closures of exponential families

The variation distance closure of an exponential family with a convex set of canonical parameters is described, assuming no regularity conditions. The tools are the concepts of convex core of a measure and extension of an exponential family, introduced previously by the authors, and a new concept of accessible faces of a convex set. Two other closures related to the information divergence are also characterized.

Piecewise linear conditional information inequality

A new information inequality of non-Shannon type is proved for three discrete random variables under conditional independence constraints, using the framework of entropy functions and polymatroids. Tightness of the inequality is described via quasi-groups

The Quantum Entropy Cone of Stabiliser States

We investigate the universal linear inequalities that hold for the von Neumann entropies in a multi-party system, prepared in a stabiliser state. We demonstrate here that entropy vectors for stabiliser states satisfy, in addition to the classic inequalities, a type of linear rank inequalities associated with the combinatorial structure of normal subgroups of certain matrix groups. In the 4-party case, there is only one such inequality, the so-called Ingleton inequality. For these systems we show that strong subadditivity, weak monotonicity and Ingleton inequality exactly characterize the entropy cone for stabiliser states.

Entropy Region and Convolution

The entropy region is constructed from vectors of random variables by collecting Shannon entropies of all subvectors. Its shape is studied here by means of polymatroidal constructions, notably by convolution. The closure of the region is decomposed into the direct sum of tight and modular parts, reducing the study to the tight part. The relative interior of the reduction belongs to the entropy region. Behavior of the decomposition under self-adhesivity is clarified. Results are specialized and extended to the region constructed from four tuples of random variables. This and computer experiments help to visualize approximations of a symmetrized part of the entropy region. The four-atom conjecture on the minimal Ingleton score is refuted.

Extreme convex set functions with many nonnegative differences

Abstract Where N is a finite set of the cardinality n and P the family of all its subsets, we study real functions on P having nonnegative differences of orders n - 2, n - 1 and n . Nonnegative differences of zeroth order, first-order, and second-order may be interpreted as nonnegativity, nonincreasingness and convexity, respectively. If all differences up to order n of a function are nonnegative, the set function is called completely monotone in analogy to the continuous case. We present a discrete Bernstein-type theorem for these functions with Mobius inversion in the place of Laplace one. Numbers of all extreme functions with nonnegative differences up to the orders n , n - 1 and n - 2, which is the most sophisticated case, and their Mobius transforms are found. As an example, we write out all extreme nonnegative nondecreasing and semimodular functions to the set N with four elements.

On minimization of multivariate entropy functionals

The problem of minimizing convex integral functionals subject to moment-like constraints is treated in a general setting when the underlying convex function may be multivariate, perhaps not strictly convex or differentiable. The results are applied to the minimization of f-divergences simultaneously in both variables.