This paper suggests that information geometry may form a natural framework to deal with the unknown part of the boundary of entropic region. An application of information geometry shows that distributions associated with Shannon facets can be associated, in the right coordinates, with affine collections of distributions. This observation allows an information geometric reinterpretation of the Shannon-type inequalities as arising from a Pythagorean style relationship. The set of distributions which violate Ingleton's inequality, and hence are linked with the part of the entropic region which is yet undetermined, is shown also to have a surprising affine information geometric structure in a special case involving four random variables and a certain support. These facts provide strong evidence for the link between information geometry and characterizing the boundary of the entropic region.
This paper suggests that information geometry may form a natural framework to deal with the unknown part of the boundary of entropic region. An application of information geometry shows that distributions associated with Shannon facets can be associated, in the right coordinates, with affine collections of distributions. This observation allows an information geometric reinterpretation of the Shannon-type inequalities as arising from a Pythagorean style relationship. The set of distributions which violate Ingleton's inequality, and hence are linked with the part of the entropic region which is yet undetermined, is shown also to have a surprising affine information geometric structure in a special case involving four random variables and a certain support. These facts provide strong evidence for the link between information geometry and characterizing the boundary of the entropic region.
Extreme point representations are utilized together with facet membership to gain insights about the relationship between various bounds on the set of normalized entropic vectors in 4 random variables. Echoing prior results from Matúš (1995), we begin by showing that the 206 extreme points of the Shannon outer bound comprise 33 forms where 16 of these forms are entropic and achievable with binary random variables (one bit), 16 of the forms are entropic and achievable using two bits, and the remaining single form of six extreme points is not entropic, not achievable with two bits, and is the only violator of the Ingleton inequality. Thus, the Ingleton inner bound is matched to the Shannon outer bound in that it contains all entropic Shannon extreme points. Further novel investigation of the structure of the Ingleton and Shannon polytopes reveal that every facet of the Shannon outer bound requires two bits per variable and can be improved upon. The inner bound formed from the convex hull of the set of binary entropic vectors is shown to neither contain nor be contained in the Ingleton inner bound, and is shown to not be a polytope. The addition of the non-Shannon-type inequality due to Zhang and Yueng plus the six non-Shannon-type inequalities due to Dougherty, Freiling, and Zeger yield a polytopic outer bound where none of the non-Shannon extreme points obey Ingleton or are binary achievable. Worst case distance gaps between the various bounds are calculated.
The entropic region is formed by the collection of the Shannon entropies of all subvectors of finitely many jointly distributed discrete random variables. For four or more variables, the structure of the entropic region is mostly unknown. We utilize a variant of the Maximum Entropy Method to obtain five-variable non-Shannon entropy inequalities, which delimit the five-variable entropy region. This method adds copies of some of the random variables in generations. A significant reduction in computational complexity, achieved through theoretical considerations and by harnessing the inherent symmetries, allowed us to calculate all five-variable non-Shannon inequalities provided by the first nine generations. Based on the results, we define two infinite collections of such inequalities and prove them to be entropy inequalities. We investigate downward-closed subsets of non-negative lattice points that parameterize these collections, and based on this, we develop an algorithm to enumerate all extremal inequalities. The discovered set of entropy inequalities is conjectured to characterize the applied method completely.
Given n (discrete or continuous) random variables Xi, the (2n- 1)-dimensional vector obtained by evaluating the joint entropy of all non-empty subsets of {X1,hellip, Xn} is called an entropic vector. Determining the region of entropic vectors is an important open problem in information theory. Recently, Chan has shown that the entropy regions for discrete and continuous random variables, though different, can be determined from one another. An important class of continuous random variables are those that are vector-valued and jointly Gaussian. It is known that Gaussian random variables violate the Ingleton bound, which many random variables such as those obtained from linear codes over finite fields do satisfy, and they also achieve certain non-Shannon inequalities. In this paper we give a full characterization of the entropy region for three jointly-Gaussian vector-valued random variables and, rather surprisingly, show that the region is strictly smaller than the entropy region for three arbitrary random variables. However, we also show the following result. For any given entropic vector h isin R7, there exists a thetas* > 0, such that for all thetas ges thetas*, the vector 1/thetas h can be generated by three vector-valued jointly Gaussian random variables. This implies that for three random variables the region of entropic vectors can be obtained by considering the cone generated by the space of Gaussian entropic vectors. It also suggests that studying Gaussian random variables for n ges 4 may be a fruitful approach to studying the space of entropic vectors for arbitrary n.
Bell's theorem in physics, as well as causal discovery in machine learning, both face the problem of deciding whether observed data is compatible with a presumed causal relationship between the variables (for example, a local hidden variable model). Traditionally, Bell inequalities have been used to describe the restrictions imposed by causal structures on marginal distributions. However, some structures give rise to non-convex constraints on the accessible data, and it has recently been noted that linear inequalities on the observable entropies capture these situations more naturally. In this paper, we show the versatility of the entropic approach by greatly expanding the set of scenarios for which entropic constraints are known. For the first time, we treat Bell scenarios involving multiple parties and multiple observables per party. Going beyond the usual Bell setup, we exhibit inequalities for scenarios with extra conditional independence assumptions, as well as a limited amount of shared randomness between the parties. Many of our results are based on a geometric observation: Bell polytopes for two-outcome measurements can be naturally imbedded into the convex cone of attainable marginal entropies. Thus, any entropic inequality can be translated into one valid for probabilities. In some situations the converse also holds, which provides us with a rich source of candidate entropic inequalities.
Causal discovery, the process of inferring cause-and-effect relationships from observational data, is a fundamental challenge across science and engineering. Traditional approaches often rely on assumptions about functional forms, linearity, or acyclicity, which may not hold in complex systems. This paper explores the "entropic geometry" framework as a powerful, non-parametric approach to causal discovery. By representing causal structures within the convex cone of attainable marginal entropies, this method can uncover intricate, non-convex constraints that are inaccessible to standard probabilistic techniques, particularly in scenarios involving hidden variables or quantum entanglement. We review the theoretical underpinnings of entropic geometry, illustrating how entropic inequalities precisely characterize the compatibility of observed data with various causal models. Furthermore, we discuss its versatility in expanding the set of known entropic constraints for complex scenarios, including multi-party interactions and systems with shared randomness. The paper highlights how this geometric perspective provides a rigorous and expressive language for understanding causal relationships, bridging concepts from information theory, statistical physics, and machine learning. We discuss the implications for empirical causal inference, computational challenges, and future research directions, positioning entropic geometry as a critical tool for advancing causal understanding in diverse scientific disciplines.
A causal structure is a relationship between observed variables that in general restricts the possible correlations between them. This relationship can be mediated by unobserved systems, modelled by random variables in the classical case or joint quantum systems in the quantum case. One way to differentiate between the correlations realisable by two different causal structures is to use entropy vectors, i.e., vectors whose components correspond to the entropies of each subset of the observed variables. To date, the starting point for deriving entropic constraints within causal structures are the so-called Shannon inequalities (positivity of entropy, conditional entropy and conditional mutual information). In the present work we investigate what happens when non-Shannon entropic inequalities are included as well. We show that in general these lead to tighter outer approximations of the set of realisable entropy vectors and hence enable a sharper distinction of different causal structures. Since non-Shannon inequalities can only be applied amongst classical variables, it might be expected that their use enables an entropic distinction between classical and quantum causal structures. However, this remains an open question. We also introduce techniques for deriving inner approximations to the allowed sets of entropy vectors for a given causal structure. These are useful for proving tightness of outer approximations or for finding interesting regions of entropy space. We illustrate these techniques in several scenarios, including the triangle causal structure.