Sampling-based proofs of almost-periodicity results and algorithmic applications

TL;DR

This paper introduces combinatorial, L^p-norm free proofs of almost-periodicity in sumsets, leading to new algorithmic versions of key results in additive combinatorics with improved efficiency and broader applicability.

Contribution

It provides an alternative proof approach for almost-periodicity, enabling the development of quasipolynomial-time algorithms for related problems in additive combinatorics.

Findings

New combinatorial proofs of almost-periodicity results

Algorithmic versions of the Bogolyubov-Ruzsa lemma and Goldreich-Levin theorem

Improved quasipolynomial algorithms for sumset problems

Abstract

We give new combinatorial proofs of known almost-periodicity results for sumsets of sets with small doubling in the spirit of Croot and Sisask, whose almost-periodicity lemma has had far-reaching implications in additive combinatorics. We provide an alternative (and L^p-norm free) point of view, which allows for proofs to easily be converted to probabilistic algorithms that decide membership in almost-periodic sumsets of dense subsets of F_2^n. As an application, we give a new algorithmic version of the quasipolynomial Bogolyubov-Ruzsa lemma recently proved by Sanders. Together with the results by the last two authors, this implies an algorithmic version of the quadratic Goldreich-Levin theorem in which the number of terms in the quadratic Fourier decomposition of a given function is quasipolynomial in the error parameter, compared with an exponential dependence previously proved by the authors. It also improves the running time of the algorithm to have quasipolynomial dependence instead of an exponential one. We also give an application to the problem of finding large subspaces in sumsets of dense sets. Green showed that the sumset of a dense subset of F_2^n contains a large subspace. Using Fourier analytic methods, Sanders proved that such a subspace must have dimension bounded below by a constant times the density times n. We provide an alternative (and L^p norm-free) proof of a comparable bound, which is analogous to a recent result of Croot, Laba and Sisask in the integers.