English

Limiting behavior of largest entry of random tensor constructed by high-dimensional data

Probability 2019-10-29 v1 Statistics Theory Statistics Theory

Abstract

Let Xk=(xk1,,xkp),k=1,,n{X}_{k}=(x_{k1}, \cdots, x_{kp})', k=1,\cdots,n, be a random sample of size nn coming from a pp-dimensional population. For a fixed integer m2m\geq 2, consider a hypercubic random tensor T\mathbf{{T}} of mm-th order and rank nn with \begin{eqnarray*} \mathbf{{T}}= \sum_{k=1}^{n}\underbrace{{X}_{k}\otimes\cdots\otimes {X}_{k}}_{m~multiple}=\Big(\sum_{k=1}^{n} x_{ki_{1}}x_{ki_{2}}\cdots x_{ki_{m}}\Big)_{1\leq i_{1},\cdots, i_{m}\leq p}. \end{eqnarray*} Let WnW_n be the largest off-diagonal entry of T\mathbf{{T}}. We derive the asymptotic distribution of WnW_n under a suitable normalization for two cases. They are the ultra-high dimension case with pp\to\infty and logp=o(nβ)\log p=o(n^{\beta}) and the high-dimension case with pp\to \infty and p=O(nα)p=O(n^{\alpha}) where α,β>0\alpha,\beta>0. The normalizing constant of WnW_n depends on mm and the limiting distribution of WnW_n is a Gumbel-type distribution involved with parameter mm.

Keywords

Cite

@article{arxiv.1910.12701,
  title  = {Limiting behavior of largest entry of random tensor constructed by high-dimensional data},
  author = {Tiefeng Jiang and Junshan Xie},
  journal= {arXiv preprint arXiv:1910.12701},
  year   = {2019}
}

Comments

21 pages

R2 v1 2026-06-23T11:57:12.890Z