Abstract
We resolve the space complexity of one-pass streaming algorithms for Minimum Dominating Set (MDS) in both insertion-only and insertion-deletion streams (up to poly-logarithmic factors) where an input graph is revealed by a sequence of edge updates. Recently, streaming algorithms for the related Set Cover problem have received significant attention. Even though MDS can be viewed as a special case of Set Cover, it is however harder to solve in the streaming setting since the input stream consists of individual edges rather than entire vertex-neighborhoods, as is the case in Set Cover.
We prove the following results (n is the number of vertices of the input graph):
1) In insertion-only streams, we give a one-pass semi-streaming algorithm (meaning Õ(n) space) with approximation factor Õ(√n). We also prove that every one-pass streaming algorithm with space o(n) has an approximation factor of Ω(n/log n).
Combined with a result by [Assadi et al., STOC'16] for Set Cover which, translated to MDS, shows that space Θ̃(n² / α) is necessary and sufficient for computing an α-approximation for every α = o(√n), this completely settles the space requirements for MDS in the insertion-only setting.
2) In insertion-deletion streams, we prove that space Ω(n² / (α log n)) is necessary for every approximation factor α ≤ Θ(n / log³ n). Combined with the Set Cover algorithm of [Assadi et al., STOC'16], which can be adapted to MDS even in the insertion-deletion setting to give an α-approximation in Õ(n² / α) space, this completely settles the space requirements for MDS in the insertion-deletion setting.
We prove the following results (n is the number of vertices of the input graph):
1) In insertion-only streams, we give a one-pass semi-streaming algorithm (meaning Õ(n) space) with approximation factor Õ(√n). We also prove that every one-pass streaming algorithm with space o(n) has an approximation factor of Ω(n/log n).
Combined with a result by [Assadi et al., STOC'16] for Set Cover which, translated to MDS, shows that space Θ̃(n² / α) is necessary and sufficient for computing an α-approximation for every α = o(√n), this completely settles the space requirements for MDS in the insertion-only setting.
2) In insertion-deletion streams, we prove that space Ω(n² / (α log n)) is necessary for every approximation factor α ≤ Θ(n / log³ n). Combined with the Set Cover algorithm of [Assadi et al., STOC'16], which can be adapted to MDS even in the insertion-deletion setting to give an α-approximation in Õ(n² / α) space, this completely settles the space requirements for MDS in the insertion-deletion setting.
| Original language | English |
|---|---|
| Title of host publication | 13th Innovations in Theoretical Computer Science Conference, ITCS 2022, January 31 - February 3, 2022, Berkeley, CA, USA |
| Editors | Mark Braverman |
| Publisher | Schloss Dagstuhl - Leibniz-Zentrum für Informatik |
| Pages | 93:1-93:23 |
| Volume | 215 |
| ISBN (Electronic) | 9783959772174 |
| DOIs | |
| Publication status | Published - 25 Jan 2022 |
Publication series
| Name | Leibniz International Proceedings in Informatics, LIPIcs |
|---|---|
| Volume | 215 |
| ISSN (Print) | 1868-8969 |
Bibliographical note
Funding Information:Supported in part by NSF awards CCF-1910534, CCF-1926872, and
Funding Information:
Supported by EPSRC New Investigator Award EP/V010611/1.
Funding Information:
Sanjeev Khanna: Supported in part by NSF awards CCF-1910534, CCF-1926872, and CCF-2045128. Christian Konrad: Supported by EPSRC New Investigator Award EP/V010611/1.
Publisher Copyright:
© Sanjeev Khanna and Christian Konrad; licensed under Creative Commons License CC-BY 4.0
Fingerprint
Dive into the research topics of 'Optimal Bounds for Dominating Set in Graph Streams'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver