English

Certifying Euclidean Sections and Finding Planted Sparse Vectors Beyond the $\sqrt{n}$ Dimension Threshold

Data Structures and Algorithms 2024-05-10 v1 Computational Complexity Metric Geometry

Abstract

We consider the task of certifying that a random dd-dimensional subspace XX in Rn\mathbb{R}^n is well-spread - every vector xXx \in X satisfies cnx2x1nx2c\sqrt{n} \|x\|_2 \leq \|x\|_1 \leq \sqrt{n}\|x\|_2. In a seminal work, Barak et. al. showed a polynomial-time certification algorithm when dO(n)d \leq O(\sqrt{n}). On the other hand, when dnd \gg \sqrt{n}, the certification task is information-theoretically possible but there is evidence that it is computationally hard [MW21,Cd22], a phenomenon known as the information-computation gap. In this paper, we give subexponential-time certification algorithms in the dnd \gg \sqrt{n} regime. Our algorithm runs in time exp(O~(nε))\exp(\widetilde{O}(n^{\varepsilon})) when dO~(n(1+ε)/2)d \leq \widetilde{O}(n^{(1+\varepsilon)/2}), establishing a smooth trade-off between runtime and the dimension. Our techniques naturally extend to the related planted problem, where the task is to recover a sparse vector planted in a random subspace. Our algorithm achieves the same runtime and dimension trade-off for this task.

Keywords

Cite

@article{arxiv.2405.05373,
  title  = {Certifying Euclidean Sections and Finding Planted Sparse Vectors Beyond the $\sqrt{n}$ Dimension Threshold},
  author = {Venkatesan Guruswami and Jun-Ting Hsieh and Prasad Raghavendra},
  journal= {arXiv preprint arXiv:2405.05373},
  year   = {2024}
}

Comments

32 pages, 2 Figures

R2 v1 2026-06-28T16:21:21.812Z