Local Algorithms for Sparsification of Average-case Graphs
Name
Cao-ruidicao-meng-eecs-2022-thesis.pdf
Description
Thesis PDF
Size
532.35 KB
Format
Adobe PDF
Checksum (MD5)
be039fbf42196b5bdd4f356e5e270617
Author(s)
Cao, Ruidi
Advisor(s)
Rubinfeld, Ronitt
Date Issued
May 2022
Publisher
Massachusetts Institute of Technology
Abstract
Given an input graph 𝐺, a Local Computation Algorithm for sparse spanning graphs provides query access to a sparse subgraph 𝐺′ ⊆ 𝐺, where 𝐺′ maintains the connectivity and/or distances in 𝐺, by making a sublinear number of probes to the input 𝐺 for each query to 𝐺′ . It is known that worst-case graphs require Ω(√ 𝑛) probes in order to detect whether a specific edge 𝑒 ∈ 𝐺′ . We want to show that, in expectation, this task can be accomplished much faster, by considering average-case graphs such as Erdos-Renyi random graphs and the Preferential Attachment model. We first present an LCA algorithm which, on an Erdos-Renyi graph input 𝐺 with edge parameter 𝑝 ≥ Ω(log(𝑛) 𝑛 ), gives fast access to a sparsification 𝐺′ of 𝐺, such that 𝐺′ is connected and has 𝑛+𝑜(𝑛) edges. Queries to 𝐺′ are answered 𝒪(∆ log2 (𝑛)) probes to 𝐺 (where ∆ = 𝒪(𝑝𝑛) is the maximum degree). We then show an LCA algorithm that, for an Erdos-Renyi graph 𝐺 with edge parameter 𝑝 ≥ Ω(log(𝑛)/√ 𝑛 ), gives access to a 4-spanner 𝐺′ of 𝐺 in 𝒪(log2/(𝑛)) probes in expectation per query, such that 𝐺′ has at most 2𝑛 edges. Finally, we give an LCA that runs on a Preferential Attachment graph 𝐺 with edge parameter Θ(log(𝑛)), which gives fast access to a sparsification 𝐺′ of 𝐺 where 𝐺′ is connected and has 𝑛 + 𝑜(𝑛) edges. Each query to 𝐺′ takes an expected 𝒪(log3 (𝑛)) probes to 𝐺.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
In Copyright - Educational Use Permitted
Copyright MIT
Persistent DSpace Link
