2026-07-24T17:50:20Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:18654 2024-03-06T11:02:22Z ddc:004 open_access

On Hashing by (Random) Equations (Invited Talk) Dietzfelbinger, Martin Hashing Retrieval Linear equations Randomness The talk will consider aspects of the following setup: Assume for each (key) x from a set 𝒰 (the universe) a vector a_x = (a_{x,0},… ,a_{x,{m-1}}) has been chosen. Given a list Z = (z_i)_{i ∈ [m]} of vectors in {0,1}^r we obtain a mapping φ_Z: 𝒰 → {0,1}^r, x ↦ ⟨a_x,Z⟩ := ⨁_{i ∈ [m]} a_{x,i} ⋅ z_i, where ⨁ is bitwise XOR. The simplest way for creating a data structure for calculating φ_Z is to store Z in an array Z[0..m-1] and answer a query for x by returning ⟨ a_x,Z⟩. The length m of the array should be (1+ε)n for some small ε, and calculating this inner product should be fast. In the focus of the talk is the case where for all or for most of the sets S ⊆ 𝒰 of a certain size n the vectors a_x, x ∈ S, are linearly independent. Choosing Z at random will lead to hash families of various degrees of independence. We will be mostly interested in the case where a_x, x ∈ 𝒰 are chosen independently at random from {0,1}^m, according to some distribution 𝒟. We wish to construct (static) retrieval data structures, which means that S ⊂ 𝒰 and some mapping f: S → {0,1}^r are given, and the task is to find Z such that the restriction of φ_Z to S is f. For creating such a data structure it is necessary to solve the linear system (a_x)_{x ∈ S} ⋅ Z = (f(x))_{x ∈ S} for Z. Two problems are central: Under what conditions on m and 𝒟 can we expect the vectors a_x, x ∈ S to be linearly independent, and how can we arrange things so that in this case the system can be solved fast, in particular in time close to linear (in n, assuming a reasonable machine model)? Solutions to these problems, some classical and others that have emerged only in recent years, will be discussed. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Martin Dietzfelbinger 2023 Is Part Of LIPIcs, Volume 274, 31st Annual European Symposium on Algorithms (ESA 2023) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/LIPIcs.ESA.2023.1 urn:nbn:de:0030-drops-186545 https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.ESA.2023.1 eng https://creativecommons.org/licenses/by/4.0/legalcode