arXiv++ Combinatorics

We study the utility and limitations of using $k$-uniform hypergraphs $H = ([n], E)$ ($n \ge \mathrm{poly}(k)$) in the context of error reduction for randomized algorithms for decision problems with one- or two-sided error. Our error reduction idea is sampling a uniformly random hyperedge of $H$, and repeating the algorithm $k$ times using the hyperedge vertices as seeds. This is a general paradigm, which captures every pseudorandom method generating $k$ seeds without repetition. We show two results which imply a gap between the typical and the worst-case behavior of using $H$ for error-reduction. First, in the context of one-sided error reduction, if using a random hyperedge of $H$ decreases the error probability from $p$ to $p^k + ε$, then $H$ cannot have too few edges, i.e., $|E| = Ω(n k^{-1} ε^{-1})$. Thus, the number of random bits needed for reducing the error from $p$ to $p^k + ε$ cannot be reduced below $\lg n+\lg(ε^{-1})-\lg k+O(1)$. This is also true for hypergraphs of average uniformity $k$. Our result implies new lower bounds for dispersers and vertex-expanders. Second, if the vertex degrees are reasonably distributed, we show that in a $(1-o(1))$-fraction of the cases, choosing $k$ pseudorandom seeds using $H$ will reduce the error probability to at most $o(1)$ above the error probability of using $k$ IID seeds, for both algorithms with one- or two-sided error. Thus, despite our lower bound, for a $(1-o(1))$-fraction of randomized algorithms (and inputs) for decision problems, the advantage of using IID samples over samples obtained from a uniformly random edge of a reasonable hypergraph is negligible. A similar statement holds true for randomized algorithms with two-sided error.

We study sheaves on posets, showing that cosystolic expansion of such sheaves can be derived from local expansion conditions of the sheaf and the poset (typically a high dimensional expander). When the poset at hand is a cell complex, a sheaf on it may be thought of as generalizing coefficient groups used for defining homology and cohomology, by letting the coefficient group vary along the cell complex. Previous works established local criteria for cosystolic expansion only for simplicial complexes and with respect to constant coefficients. Cosystolic expansion of sheaves is related to property testing. We use this relation and our local criterion for cosystolic expansion to give two applications to locally testable codes (LTCs). First, we show the existence of good $2$-query LTCs. These codes are related to the recent good $q$-query LTCs of Dinur et. al and Panteleev-Kalachev, being the formers' so-called line codes, but we get them from a new, more illuminating perspective, namely, by realizing them as cocycle codes of sheaves over posets. We then derive their good properties directly from our criterion for cosystolic expansion. Second, we give a local criterion for a a lifted code (with some auxiliary structure) to be locally testable. This improves on a previous work of Dikstein et. al, where it was shown that one can obtain local testability of lifted codes from a mixture of local and global conditions.

The Cheeger constant of a graph, or equivalently its coboundary expansion, quantifies the expansion of the graph. This notion assumes an implicit choice of a coefficient group, namely, $\mathbb{F}_2$. In this paper, we study Cheeger-type inequalities for graphs endowed with a generalized coefficient group, called a sheaf; this is motivated by applications to cosystolic expansion and locally testable codes. We prove that a graph is a good spectral expander if and only if it has good coboundary expansion relative to any (resp. some) constant sheaf, or equivalently, relative to any `ordinary' coefficient group. We moreover show that sheaves that are close to being constant in a well-defined sense are also good coboundary expanders, provided that their underlying graph is an expander, thus giving the first example of good coboundary expansion in non-cosntant sheaves on sparse graphs. By contrast, we observe that for general sheaves on graphs, it is impossible to relate the expansion of the graph and the coboundary expansion of the sheaf. We specialize our results to sheaves on (finite) spherical buildings. Specifically, we show that the normalized second eigenvalue of the (weighted) graph underlying a $q$-thick $d$-dimensional spherical building is $O(\frac{1}{\sqrt{q}-3d})$ if $q>9d^2$. Plugging this into our results about coboundary expansion gives explicit lower bounds on the coboundary expansion of some constant and non-constant sheaves on spherical buildings; for a fixed dimension $d$, the bounds approach a constant as the thickness $q$ grows. Along the way, we prove a new version of the Expander Mixing Lemma for $r$-partite weighted graphs.

We expose a strong connection between good $2$-query locally testable codes (LTCs) and high dimensional expanders. Here, an LTC is called good if it has constant rate and linear distance. Our emphasis in this work is on LTCs testable with only $2$ queries, which are of particular interest to theoretical computer science. This is done by introducing a new object called a sheaf that is put on top of a high dimensional expander. Sheaves are vastly studied in topology. Here, we introduce sheaves on simplicial complexes. Moreover, we define a notion of an expanding sheaf that has not been studied before. We present a framework to get good infinite families of $2$-query LTCs from expanding sheaves on high dimensional expanders, utilizing towers of coverings of these high dimensional expanders. Starting with a high dimensional expander and an expanding sheaf, our framework produces an infinite family of codes admitting a $2$-query tester. We show that if the initial sheaved high dimensional expander satisfies some conditions, which can be checked in constant time, then these codes form a family of good $2$-query LTCs. We give candidates for sheaved high dimensional expanders which can be fed into our framework, in the form of an iterative process which conjecturally produces such candidates given a high dimensional expander and a special auxiliary sheaf. (We could not verify the prerequisites of our framework for these candidates directly because of computational limitations.) We analyse this process experimentally and heuristically, and identify some properties of the fundamental group of the high dimensional expander at hand which are sufficient (but not necessary) to get the desired sheaf, and consequently an infinite family of good $2$-query LTCs.

This paper brings the main definitions and results from "The Ramanujan Property for Simplicial Complexes" [arXiv:1605.02664]. No proofs are given. Given a simplicial complex $\mathcal{X}$ and a group $G$ acting on $\mathcal{X}$, we define Ramanujan quotients of $\mathcal{X}$. For $G$ and $\mathcal{X}$ suitably chosen this recovers Ramanujan $k$-regular graphs and Ramanujan complexes in the sense of Lubotzky, Samuels and Vishne. Deep results in automorphic representations are used to give new examples of Ramanujan quotients when $\mathcal{X}$ is the affine building of an inner form of $\mathbf{GL}_n$ over a local field of positive characteristic.

Let $G$ be a topological group acting on a simplicial complex $\mathcal{X}$ satisfying some mild assumptions. For example, consider a $k$-regular tree and its automorphism group, or more generally, a regular affine Bruhat-Tits building and its automorphism group. We define and study various types of high-dimensional spectra of quotients of $\mathcal{X}$ by subgroups of $G$. These spectra include the spectrum of many natural operators associated with the quotients, e.g. the high-dimensional Laplacians. We prove a theorem in the spirit of the Alon-Boppana Theorem, leading to a notion of Ramanujan quotients of $\mathcal{X}$. Ramanujan $k$-regular graphs and Ramanuajn complexes in the sense of Lubotzky, Samuels and Vishne are Ramanujan in dimension $0$ according to our definition (for $\mathcal{X}$, $G$ suitably chosen). We give a criterion for a quotient of $\mathcal{X}$ to be Ramanujan which is phrased in terms of representations of $G$, and use it, together with deep results about automorphic representations, to show that affine buildings of inner forms of $\mathbf{GL}_n$ over local fields of positive characteristic admit infinitely many quotients which are Ramanujan in all dimensions. The Ramanujan (in dimension $0$) complexes constructed by Lubotzky, Samuels and Vishne arise as a special case of our construction. Our construction also gives rise to Ramanujan graphs which are apparently new. Other applications are also discussed. For example, we show that there are non-isomorphic simiplicial complexes which are isospectral in all dimensions.