Unique-Neighbor-Like Expansion and Group-Independent Cosystolic Expansion (2211.09482v1)

Abstract: In recent years, high dimensional expanders have been found to have a variety of applications in theoretical computer science, such as efficient CSPs approximations, improved sampling and list-decoding algorithms, and more. Within that, an important high dimensional expansion notion is \emph{cosystolic expansion}, which has found applications in the construction of efficiently decodable quantum codes and in proving lower bounds for CSPs. Cosystolic expansion is considered with systems of equations over a group where the variables and equations correspond to faces of the complex. Previous works that studied cosystolic expansion were tailored to the specific group $\mathbb{F}_2$. In particular, Kaufman, Kazhdan and Lubotzky (FOCS 2014), and Evra and Kaufman (STOC 2016) in their breakthrough works, who solved a famous open question of Gromov, have studied a notion which we term ``parity'' expansion for small sets. They showed that small sets of $k$-faces have proportionally many $(k+1)$-faces that contain \emph{an odd number} of $k$-faces from the set. Parity expansion for small sets could be used to imply cosystolic expansion only over $\mathbb{F}_2$. In this work we introduce a stronger \emph{unique-neighbor-like} expansion for small sets. We show that small sets of $k$-faces have proportionally many $(k+1)$-faces that contain \emph{exactly one} $k$-face from the set. This notion is fundamentally stronger than parity expansion and cannot be implied by previous works. We then show, utilizing the new unique-neighbor-like expansion notion introduced in this work, that cosystolic expansion can be made \emph{group-independent}, i.e., unique-neighbor-like expansion for small sets implies cosystolic expansion \emph{over any group}.