← All problems

Maximum diagonal product conjecture for bounded-rank stochastic matrices

Lavi's conjecture: Let 1≤k≤n1\le k\le n, write n=qk+sn=qk+s with integers q≥1q\ge 1 and 0≤s<k0\le s<k, and let A=(aij)∈R≥0n×nA=(a_{ij})\in\mathbb{R}_{\ge 0}^{n\times n} be either row-stochastic, meaning ∑j=1naij=1\sum_{j=1}^n a_{ij}=1 for every ii, or column-stochastic, meaning ∑i=1naij=1\sum_{i=1}^n a_{ij}=1 for every jj. If rank⁡(A)≤k\operatorname{rank}(A)\le k, then for every permutation σ\sigma of {1,…,n}\{1,\ldots,n\}, ∏i=1nai,σ(i)≤q−q(k−s)(q+1)−(q+1)s\prod_{i=1}^n a_{i,\sigma(i)}\le q^{-q(k-s)}(q+1)^{-(q+1)s}. The problem also asks for a complete classification of all matrices and permutations for which equality holds.

References

Progress summary

Refreshed
Open · solution reported

A 2026 preprint claims to prove the conjecture and classify all equality cases, but the claim has not been independently checked.

The conjecture asks for a sharp upper bound on every diagonal product of a row- or column-stochastic matrix of rank at most kk, where n=qk+sn=qk+s, together with a classification of all equality cases. Yair Lavi formulated it in 2018 as Conjecture 4.

Known results

  • Lavi (2018): stated the bound (q−q)k−s((q+1)−(q+1))s\left(q^{-q}\right)^{k-s}\left((q+1)^{-(q+1)}\right)^s and the conjectured extremizers, but did not prove them.

2026 preprint claim

A 2026 preprint by Lavi reports a proof of the diagonal-product bound for all such matrices and a complete equality classification. The claim is unrefereed and has no independent mathematical assessment in the retrieved sources. A separate 2026 preprint proves only the analogous permanent statement, not this conjecture.

Current status (as of November 2026): A 2026 preprint claims the general bound and equality classification, while the 2018 conjecture remains unverified because no independent assessment was found.

Sources

Solutions 0

No solutions have been posted yet.