🔢 mathematics

Revisiting the Stability of the Ingleton Inequality: A Tropicalization-Free Approach

This paper revisits the stability of the Ingleton inequality under small conditional independence violations by replacing the complex tropical probability framework with a streamlined, tropicalization-free approach that yields explicit error terms, improved estimates, and a resolution to an open problem regarding the stability of the sum of two Ingleton expressions.

Original authors: Laszlo Csirmaz

Published 2026-07-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Laszlo Csirmaz

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to map out the shape of a mysterious, invisible landscape. This isn't a landscape of mountains and rivers, but of information. In the world of information theory, scientists study how much "stuff" (data) is contained in different groups of things and how they relate to one another. To do this, they use a mathematical ruler called Shannon entropy, which measures the amount of surprise or uncertainty in a set of data. If you have a deck of cards, the entropy is high because you don't know what the next card will be; if you have a deck of all Aces, the entropy is zero because there is no surprise.

Now, imagine you have four friends sharing secrets. There are strict rules about how their secrets can overlap. One of the most famous rules is called the Ingleton inequality. Think of this rule as a "law of the land" that says, "In a perfectly logical world, the way these four friends share secrets must follow a specific balance." For a long time, mathematicians knew this rule held true for certain types of logical structures (like representable matroids), but they weren't sure what happened if the world wasn't perfect. What if the friends' secrets were slightly messy, or if the rules of independence were only almost true? This is where the question of stability comes in: if you nudge the system just a tiny bit, does the rule break completely, or does it just bend a little? Understanding this helps us know the true shape of the "entropy landscape" and whether our mathematical models of information are robust or fragile.

The Paper's Story: Smoothing the Rough Edges

In a recent study, researchers Matveev and Romashchenko investigated exactly this question: how much can the Ingleton inequality be broken if the conditions are only almost met? They used a very complex, high-tech tool called "tropical probability spaces" to solve it. Think of tropical probability like looking at a landscape through a special pair of glasses that turns everything into a grid of sharp, angular lines. It works, but it's complicated and hard to see the smooth curves underneath.

The author of this paper, Laszlo Csirmaz, decided to take off those special glasses. He wanted to see the landscape with his own eyes, using traditional, "tropicalization-free" math. His goal was to prove the same results but with a simpler, clearer approach that didn't rely on that complex grid framework.

What the paper found:
The author successfully recreated the stability results using this simpler, traditional method. He proved that even when the conditions for the Ingleton inequality are slightly off (meaning the "conditional mutual information" terms are small but not zero), the inequality doesn't collapse entirely. Instead, it bends in a predictable way.

Crucially, the paper does three main things:

It simplifies the math: By avoiding the complex "tropical" framework, the proofs become much shorter and easier to follow.
It gives exact numbers: Previous work often used vague "big O" notation (which just says "it gets small eventually"). This paper provides explicit error terms. It tells you exactly how much the inequality can be violated based on how small the initial error is. For example, if the error in the conditions is $1/k^2$ , the violation of the inequality is bounded by something like $3/k$ .
It solves a mystery: The previous researchers asked if there was a specific family of rules (inequalities) that could explain the stability of the sum of two Ingleton expressions. The author says, "Yes, there is!" He discovered and proved a new, infinite family of entropy inequalities that settles this question.

What the paper rules out:
The paper explicitly argues against the idea that you need the complex tropical probability framework to understand these stability results. It shows that the "special glasses" aren't necessary; you can get the same (and sometimes better) results with standard tools.

How sure are we?
The author is very confident. These are not simulations or guesses; they are mathematical proofs. The paper provides rigorous, step-by-step logical arguments (using tools like the "Copy Lemma" and "Ahlswede–Körner reduction") to demonstrate that the new inequalities hold true.

The "Corner" of the Landscape:
The paper also points out a subtle difference in the landscape. In one scenario (Theorem 8), the "bending" of the rule follows a square-root pattern (like $\sqrt{\epsilon}$ ), which is a bit "rougher." In another scenario where more symmetry is present (Theorem 10), the bending is much smoother, almost linear (like $\epsilon \log(1/\epsilon)$ ). The author suggests that maybe the "rough" scenario could also be smoothed out, but that remains an open question for future explorers. He even offers a candidate formula (inequality 4) that, if proven, would show that the roughness can be tamed, but he admits this hasn't been proven yet—it's just a promising lead.

In short, this paper is like a cartographer who redrew a famous map. They didn't change the territory, but they removed the confusing, jagged grid lines and replaced them with smooth, clear contours, giving us exact measurements of how the land slopes when we get close to the edges. And best of all, they found a new trail (the infinite family of inequalities) that no one knew existed before.

Technical Summary: Revisiting the Stability of the Ingleton Inequality

Problem Statement
The Ingleton inequality is a fundamental constraint in the geometry of the four-variable entropy region ( $\Gamma^*_4$ ), serving as a necessary condition for representable matroids. While the inequality holds for entropic points under exact conditional independence constraints, recent work by Matveev and Romashchenko (2026) investigated the stability of these implications. Specifically, they quantified the extent to which the Ingleton inequality can be violated when a group of conditional mutual information terms is small but non-zero. Their proofs relied on the complex framework of "tropical probability spaces," a tool designed to preserve subtle properties of probability spaces often lost in traditional entropy profiles. The current paper addresses the need for a more accessible framework to analyze these stability results, aiming to derive explicit error terms and resolve open questions regarding infinite families of entropy inequalities.

Methodology
The paper proposes a "tropicalization-free" approach, utilizing traditional Information Theory and the geometry of polymatroids. The core methodology involves three main technical components:

The Copy Lemma and Symmetry: The authors utilize the Copy Lemma (Lemma 1) to construct extensions of almost entropic (aent) functions. This involves adding a "copy" of a variable $a$ over a set $X$ such that the copy is conditionally independent of other variables given $X$ . This construction preserves symmetry between variables, a crucial property for the subsequent proofs.
Iterative Construction (Construction 1 & 2): The authors develop a recursive construction that generates a sequence of variables $z_0, \dots, z_k$ $z_{0}, \dots, z_{k}$ (and subsequently $w_0, \dots, w_k$ $w_{0}, \dots, w_{k}$ ) with specific entropy properties.
- Construction 1 creates variables where the entropy of the new variable conditioned on a set $S$ scales linearly with the index $i$ relative to the conditional entropy of the original variable.
- Construction 2 combines Construction 1 with the Ahlswede–Körner reduction (Lemma 6). This reduction maps an aent function to a new aent function where a specific variable $w$ has zero entropy conditioned on a base set, effectively "gluing" distributions to enforce conditional independence while maintaining the aent property.
Explicit Error Analysis: Unlike previous works that relied on asymptotic $O(\cdot)$ notation, this approach derives explicit, non-asymptotic error terms. The proofs involve summing Shannon inequalities over the constructed sequence of variables and bounding the cumulative error using the initial smallness of the conditional mutual information terms.

Key Contributions and Results

Streamlined Proofs: The paper successfully reproduces the stability results of Matveev and Romashchenko (2026) without using tropical probability spaces. This simplifies the underlying concepts and makes the proofs more accessible to the broader information theory community.
Explicit Error Terms: The authors provide explicit bounds for the violation of the Ingleton inequality.
- Theorem 8: Establishes a lower bound for the Ingleton expression when $I(x, a|y) + I(y, a|x)$ is small (specifically $\le (1/k^2)I(x, y)$ ). The violation is bounded by $-(3/k)I(x, y)$ .
- Theorem 10: Under stronger symmetry conditions (smallness of both $a$ and $b$ terms), the paper derives an "almost linear" error bound of the form $-(3k/2^k)I(x, y)$ , which is significantly tighter than the bound in Theorem 8.
Resolution of an Open Problem: The paper resolves an open question posed in prior work regarding the existence of an infinite family of entropy inequalities that establishes the stability of the sum of two Ingleton expressions.
- Theorem 11: The authors exhibit a new infinite family of entropy inequalities (parameterized by integer $k \ge 1$ ) that proves the stability of $ing(x, y, a, b) + ing(x, y, c, d)$. This family explicitly relates the sum of the Ingleton expressions to the conditional mutual information terms $I(x, a|y)$ and $I(x, b|y)$ .
Refined Estimates: The paper improves upon previous estimates by relaxing certain conditions to strengthen the bounds, particularly in the transition from square-root error terms to almost linear ones under specific symmetry assumptions.

Significance
The paper claims significance primarily in its methodological shift and the precision of its results. By bypassing the machinery of tropical probability spaces, the authors demonstrate that the stability of the Ingleton inequality can be understood through traditional, "classical" information-theoretic tools. This approach not only streamlines the proofs but also yields explicit, non-asymptotic error terms, which are more practical for quantitative analysis than asymptotic bounds. Furthermore, by providing a new infinite family of inequalities (Theorem 11), the paper settles a specific open problem regarding the joint stability of multiple Ingleton expressions.

The authors note that while Theorem 9 (derived from Theorem 11) currently exhibits a square-root error term ( $O(\sqrt{\epsilon})$ ), Theorem 10 suggests that an almost linear bound ( $O(\epsilon \log(1/\epsilon))$ ) is possible under stronger symmetry. The paper concludes by identifying the formal proof of a candidate inequality (Equation 4) that would unify these results as a promising direction for future research, without claiming to have solved it yet.

The Paper's Story: Smoothing the Rough Edges

Technical Summary: Revisiting the Stability of the Ingleton Inequality

More like this