Reformulation of the Covering and Quantizer Problems as Ground States of Interacting Particles

TL;DR

This paper reformulates the covering and quantizer problems as ground state energy minimization of interacting particles, enabling new analytical and numerical approaches with applications in communications, data analysis, and physics.

Contribution

It introduces a novel formulation linking geometric problems to particle ground states, allowing the use of physics-based optimization techniques.

Findings

Disordered sphere packings can provide efficient coverings in high dimensions.

Improved upper bounds on quantizer errors are derived from sphere-packing solutions.

Disordered packings can outperform lattice structures in certain dimensions.

Abstract

We reformulate the covering and quantizer problems as the determination of the ground states of interacting particles in $\mathbb{R}^d$ that generally involve single-body, two-body, three-body, and higher-body interactions. This is done by linking the covering and quantizer problems to certain optimization problems involving the "void" nearest-neighbor functions that arise in the theory of random media and statistical mechanics. These reformulations, which again exemplifies the deep interplay between geometry and physics, allow one now to employ theoretical and numerical optimization techniques to analyze and solve these energy minimization problems. The covering and quantizer problems have relevance in numerous applications, including wireless communication network layouts, the search of high-dimensional data parameter spaces, stereotactic radiation therapy, data compression, digital communications, meshing of space for numerical analysis, and coding and cryptography, among other examples. The connections between the covering and quantizer problems and the sphere-packing and number-variance problems are discussed. We also show that disordered saturated sphere packings provide relatively thin (economical) coverings and may yield thinner coverings than the best known lattice coverings in sufficiently large dimensions. In the case of the quantizer problem, we derive improved upper bounds on the quantizer error using sphere-packing solutions, which are generally substantially sharper than an existing upper bound in low to moderately large dimensions. We also demonstrate that disordered saturated sphere packings yield relatively good quantizers. Finally, we remark on possible applications of our results for the detection of gravitational waves.