发表机构
University of California, Irvine(加州大学尔湾分校)
机构由 AI 辅助整理,请以论文原文为准。AI 中文总结
本文针对随机张量特征向量的样本协方差,确定了其在Marchenko–Pastur阈值处的临界谱律,推导了经验谱分布的极限并验证了均匀二次型集中的精确收敛范围。
AI 中文摘要
设$X$为均值为0、方差为1且具有有限四阶矩的随机变量,构造由$n$个独立同分布的$X$副本中所有无平方因子单项式构成的$d$阶主张量特征向量。对于$m$个独立样本,我们在临界尺度$d^2/n\to\lambda\in[0,\infty)$及长宽比$p/m\to c$的范围内确定样本协方差的全局谱律。对于具有有限四阶矩且$P(|X|=1)<1$的固定基础分布,现有研究表明当且仅当$d=o(\sqrt{n})$时,会出现普通Marchenko–Pastur收敛。我们确定了有限临界边界:当$d^2/n\to\lambda\in(0,\infty)$时,张量半径在二次Wasserstein距离下收敛于由四阶矩决定的对数正态分布,而所有剩余有界二次波动均消失。留一法预解式论证进而得到经验谱分布几乎必然收敛于由该内生对数正态跳跃驱动的自由复合泊松律。当四阶矩 excess 或重叠强度消失时,该极限退化为Marchenko–Pastur。在单位模情形下,我们的估计恢复了均匀二次型集中的精确范围$\min(d,n-d)=o(n)$,并在该整个范围内蕴含Marchenko–Pastur收敛,且带有显式方差界。
英文摘要
For a centered, variance-one random variable $X$ with finite fourth moment, let $x$ be the vector of square-free degree-$d$ monomials in $n$ independent copies of $X$. At the critical scale $d^2/n \to λ\in (0,\infty)$, the normalized squared length of $x$ converges to the lognormal variable $R = \exp(\sqrt{λv}\, Z - λv/2)$, where $v = \mathbb{E} X^4 - 1$ and $Z$ is standard normal. If $\binom{n}{d}/N \to c \in (0,\infty)$, the sample covariance of $N$ independent copies of $x$ has an almost-sure limiting spectral law: the free compound-Poisson law with rate $1/c$ and jump distribution $\operatorname{Law}(cR)$. It reduces to Marchenko-Pastur when $λv = 0$. The proof shows that subtracting the contribution of the sample length leaves vanishing quadratic-form fluctuations, even when $\mathbb{E} X^3 \neq 0$; length and direction need not be independent.