A Z2-Topological Framework for Sign-rank Lower Bounds
Abstract
We develop a topological framework for proving lower bounds on sign-rank via Z2-equivariant topology, and use it to resolve the sign-rank of the Gap Hamming Distance problem up to lower-order terms. For every (partial) sign matrix A, we associate a free Z2-simplicial complex S(A) and show that sign-rank of A is characterized by the linear analog of Z2-index of S(A). As a consequence, the classical Z2-index of S(A) lower bounds the sign-rank of A, which reduces sign-rank lower bounds to topological obstructions. This reduction allows us to use various tools from Z2-equivariant topology, particularly in regimes where classical lower-bound techniques break down. As the main application, we consider the Gap Hamming Distance function GHDkn (defined for k < n/2), which distinguishes pairs of strings in \0,1\n with Hamming distance at most k from pairs with distance at least n-k. We prove an essentially tight lower bound and show that for any k, \[ sign-rank(GHDkn) = (1-ok(1)) 2k. \] where the ok(1) term is O( kk). This improves on the previous lower bound of Hatami, Hosseini, and Meng (STOC 2023) who proved that sign-rank of GHDkn is at least (k/(n/k)). A key technical ingredient is a new analysis of the Z2-coindex (which lower bounds Z2-index) of the Vietoris-Rips complex of the hypercube in the sparse regime which yields an essentially tight lower bound. Previously, no results were known in the sparse regime.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.