clopen.solutions
OP-00214 Solution claimed

A Seginer theorem for arbitrary independent identically distributed tensor entries

Submit a result

For integers $r \ge 3$ and $n_1, \dots, n_r \ge 2$, let $T \in \mathbb{R}^{n_1 \times \cdots \times n_r}$ have independent identically distributed integrable real entries of mean zero. The distribution may depend on the dimensions, and no moment beyond the first is assumed. Let $$\|T\|_{\mathrm{inj}} = \sup_{\|x_1\|_2 = \cdots = \|x_r\|_2 = 1} |\langle T, x_1 \otimes \cdots \otimes x_r \rangle|$$ be its injective norm, and let $$F(T) = \max_{1 \le k \le r} \mathbb{E} \max_{i_j,\ j \ne k} \Big( \sum_{a=1}^{n_k} T_{i_1, \dots, i_{k-1}, a, i_{k+1}, \dots, i_r}^2 \Big)^{1/2}$$ be the largest expected Euclidean fiber norm, the maximum over the modes $k$ being taken outside the expectation. For matrices ($r = 2$) the comparison below is Seginer's theorem. Lucca and Pesenti proved it for entries distributed as a random variable $Z$ with $\mathbb{E}|Z|^{2+\varepsilon} \le K (\mathbb{E} Z^2)^{1 + \varepsilon/2}$ for some $\varepsilon \in (0, 1)$ and $K \ge 1$, with constants depending on $r$, $\varepsilon$ and $K$.

Problem. For each $r \ge 3$, are there constants $0 < c_r \le C_r < \infty$, depending only on $r$, such that $$c_r F(T) \le \mathbb{E} \|T\|_{\mathrm{inj}} \le C_r F(T)$$ for all $n_1, \dots, n_r \ge 2$ and every such entry distribution?

Source: Lucca and Pesenti (2026), Conjecture 1.8, for every order r; stated for r ≥ 3 and n_j ≥ 2 by OpenProblemsInNLA, the case r = 2 being Seginer's theorem.

Imported from github.com/ajt60gaibb/OpenProblemsInNLA.

Classification

Primary
PR — Probability
Secondary
FA, NA
Keywords
injective norms, OpenProblemsInNLA, random tensors
Literature checked
Oct. 8, 2026 by clemens-admin

References

@misc{lucca2026,
  author        = {Lucca, Kevin and Pesenti, Lucas},
  title         = {Norm Bounds for Sparse Random Tensors and Spectral Gap of Random Hypergraphs},
  year          = {2026},
  eprint        = {2607.07308},
  archivePrefix = {arXiv},
}
@article{seginer2000,
  author  = {Seginer, Yoav},
  title   = {The Expected Norm of Random Matrices},
  journal = {Combin. Probab. Comput.},
  volume  = {9},
  number  = {2},
  year    = {2000},
  pages   = {149--166},
  doi     = {10.1017/S096354830000420X},
}
@misc{openproblemsinnla-tr21,
  author       = {Townsend, Alex},
  title        = {{TR-21}: A Seginer theorem for arbitrary independent identically distributed tensor entries},
  howpublished = {Open Problems in Numerical Linear Algebra, GitHub repository},
  year         = {2026},
  url          = {https://github.com/ajt60gaibb/OpenProblemsInNLA/blob/main/tensor-computations/TR-21/README.md},
  note         = {Read at commit eb7c9d6, 2026-10-07},
}

Resolution claims

Anonymous

For every fixed order $k \ge 3$, we prove that the expected injective norm of a tensor with iid mean-zero integrable entries is comparable to its largest expected Euclidean fiber norm, uniformly over all rectangular dimensions and entry laws. Precisely, writing $F(T)$ for that fiber norm, $F(T) \le \mathbb{E}\|T\|_{\mathrm{inj}} \le C_k F(T)$, where $C_k$ depends only on $k$; no finite-variance …

Full resolution · not verified · fidelity not established · posted on GitHub 30 Sep 2026

Status history

WhenChangeByReason
2026-10-08 Needs verification → Solution claimed clemens-admin Listed with the record of its solution, unreviewed.

Citing this problem

This identifier is permanent. It will not change if the problem is solved, reclassified or withdrawn.

OP-00214, "A Seginer theorem for arbitrary independent identically distributed tensor entries", clopen.solutions.