Constant-Factor Hardness of Lattice Covering and Statistical Sample-Size Exponents
Year: 2026
Type: preprint
Abstract: This preprint studies the complexity of approximating lattice covering radii and the sample requirements for estimating collision probabilities under strong noise. The two cosets of D8 in E8 yield an exact identity relating the squared Euclidean covering radius to the Hamming covering radius of a binary linear code. A quotient construction proves constant-factor hardness for each fixed rational finite p at least one and for power-sum metrics below one. For distributions supported on prescribed lattice cosets, the critical Fourier band gives the exact high-noise sample-size exponent under Gaussian and symmetric stable noise. The paper computes the critical band explicitly in one dimension, separates dual-lattice covering from integer generation when the origin is unknown, and gives hardness results for sparse and fixed Gaussian correlations.
Source: Figshare
Author: Akihiro Koide
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7233859718
