Abstract:Covering numbers are a powerful tool used in the development of approximation algorithms, randomized dimension reduction methods, smoothed complexity analysis, and others. In this paper we prove upper bounds on the covering number of numerous sets in Euclidean space, namely real algebraic varieties, images of polynomial maps and semialgebraic sets in terms of the number of variables and degrees of the polynomials involved. The bounds remarkably improve the best known general bound by Yomdin-Comte, and our proof is much more straightforward. In particular, our result gives new bounds on the volume of the tubular neighborhood of the image of a polynomial map and a semialgebraic set, where results for varieties by Lotz and Basu-Lerario are not directly applicable. We illustrate the power of the result on three computational applications. Firstly, we derive a near-optimal bound on the covering number of low rank CP tensors, quantifying their approximation properties and filling in an important missing piece of theory for tensor dimension reduction and reconstruction. Secondly, we prove a bound on the required dimension for the randomized sketching of polynomial optimization problems, which controls how much computation can be saved through randomization without sacrificing solution quality. Finally, we deduce generalization error bounds for deep neural networks with rational or ReLU activation functions, improving or matching the best known results in the machine learning literature while helping to quantify the impact of architecture choice on generalization error.

Covering shrinking polynomials by quasi progressions

A Quantitative Bound For Szemerédi's Theorem for a Complexity One Polynomial Progression over $\mathbb{Z}/N\mathbb{Z}$

The Borwein Conjectures over Arithmetic Progressions

Covering Number of Real Algebraic Varieties and Beyond: Improved Bounds and Applications

Shrinking targets versus recurrence: the quantitative theory

Arithmetic progressions in polynomial orbits

On product sets of arithmetic progressions

Quantitative asymptotics for polynomial patterns in the primes

Improved Bounds for Szemerédi's Theorem

Arithmetic progressions of squares, cubes and $n$-th powers

On Erdős covering systems in global function fields

Prime polynomials in short intervals and in arithmetic progressions

New Bounds on Quotient Polynomials with Applications to Exact Divisibility and Divisibility Testing of Sparse Polynomials

Primes in short arithmetic progressions

Expanding polynomials for sets with additive structure

Primes in arithmetic progressions on average II

The polynomial multidimensional Szemerédi Theorem along shifted primes

A lower bound for the variance in arithmetic progressions of some multiplicative functions close to $1$

Quantitative bounds in the nonlinear Roth theorem

On converses to the polynomial method

On primes in arithmetic progressions and bounded gaps between many primes