We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computational Complexity

Title: (Gap/S)ETH Hardness of SVP

Abstract: $ \newcommand{\problem}[1]{\ensuremath{\mathrm{#1}} } \newcommand{\SVP}{\problem{SVP}} \newcommand{\ensuremath}[1]{#1} $We prove the following quantitative hardness results for the Shortest Vector Problem in the $\ell_p$ norm ($\SVP_p$), where $n$ is the rank of the input lattice.
$\bullet$ For "almost all" $p > p_0 \approx 2.1397$, there no $2^{n/C_p}$-time algorithm for $\SVP_p$ for some explicit constant $C_p > 0$ unless the (randomized) Strong Exponential Time Hypothesis (SETH) is false.
$\bullet$ For any $p > 2$, there is no $2^{o(n)}$-time algorithm for $\SVP_p$ unless the (randomized) Gap-Exponential Time Hypothesis (Gap-ETH) is false. Furthermore, for each $p > 2$, there exists a constant $\gamma_p > 1$ such that the same result holds even for $\gamma_p$-approximate $\SVP_p$.
$\bullet$ There is no $2^{o(n)}$-time algorithm for $\SVP_p$ for any $1 \leq p \leq 2$ unless either (1) (non-uniform) Gap-ETH is false; or (2) there is no family of lattices with exponential kissing number in the $\ell_2$ norm. Furthermore, for each $1 \leq p \leq 2$, there exists a constant $\gamma_p > 1$ such that the same result holds even for $\gamma_p$-approximate $\SVP_p$.
Subjects: Computational Complexity (cs.CC)
Journal reference: STOC 2018
DOI: 10.1145/3188745.3188840
Cite as: arXiv:1712.00942 [cs.CC]
  (or arXiv:1712.00942v1 [cs.CC] for this version)

Submission history

From: Noah Stephens-Davidowitz [view email]
[v1] Mon, 4 Dec 2017 07:50:40 GMT (60kb,D)

Link back to: arXiv, form interface, contact.