Skip to content

Talagrands Convexity Conjecture #1853

@franzhusch

Description

@franzhusch

What is the conjecture

A compact set $A \subseteq \mathbb{R}^N$ is balanced if whenever $x \in A$ and $\lambda \in \mathbb{R}$ with $|\lambda| \leq 1$, we have $\lambda x \in A$. Let $\gamma_N$ denote the standard Gaussian measure on $\mathbb{R}^N$.

Talagrand's Convexity Problem asks: Does there exist an integer $q$ (independent of the dimension $N$) such that for all $N$ and every compact balanced set $A \subseteq \mathbb{R}^N$ with $\gamma_N(A) \geq 1/2$, one can find a convex compact set $C \subseteq A + \cdots + A$ ($q$ terms in the Minkowski sum) such that $\gamma_N(C) \geq 1/2$?"

In other words, can one always create a convex set of high Gaussian measure by taking a finite (dimension-independent) number of Minkowski sums of a balanced set with high Gaussian measure?

(This description may contain subtle errors especially on more complex problems; for exact details, refer to the sources.)

A proof is worth 1000$ (I think, see sources for details).

Sources:

Prerequisites needed

Formalizability Rating: 1/5 (0 is best) (as of 2026-01-22)

Building blocks (from Mathlib):

  • Metric.Compact for compact sets in ℝ^N
  • MeasureTheory.gaussianMeasure for the Gaussian measure γ_N
  • Convex for convexity
  • Minkowski sum operations (basic algebraic structures)

Missing pieces:

  • Formal definition of balanced set (simple: ∀ x ∈ A, ∀ λ ∈ ℝ, |λ| ≤ 1 → λ • x ∈ A)
  • Wrapper for the universal quantifier over dimensions and q-fold Minkowski sums

Rating justification: All essential building blocks (compact sets, Gaussian measure, convex sets, Minkowski sums) exist in Mathlib. The definition of balanced sets is straightforward. The statement itself is expressible using existing infrastructure with only minor helper definitions needed.

AMS categories

  • ams-52
  • ams-28
  • ams-46

Choose either option

  • I plan on adding this conjecture to the repository
  • This issue is up for grabs: I would like to see this conjecture added by somebody else

This issue was generated by an AI agent and reviewed by me.

See more information here: link

Feedback on mistakes/hallucinations: link

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions