eng
Schloss Dagstuhl – Leibniz-Zentrum für Informatik
Leibniz International Proceedings in Informatics
1868-8969
2024-01-18
16:1
16:20
10.4230/LIPIcs.OPODIS.2023.16
article
Improved Deterministic Distributed Maximum Weight Independent Set Approximation in Sparse Graphs
Gil, Yuval
1
https://orcid.org/0009-0007-7762-3029
Technion - Israel Institute of Technology, Haifa, Israel
We design new deterministic CONGEST approximation algorithms for maximum weight independent set (MWIS) in sparse graphs. As our main results, we obtain new Δ(1+ε)-approximation algorithms as well as algorithms whose approximation ratio depend strictly on α, in graphs with maximum degree Δ and arboricity α. For (deterministic) Δ(1+ε)-approximation, the current state-of-the-art is due to a recent breakthrough by Faour et al. [SODA 2023] that showed an O(log² (Δ W)⋅ log (1/ε)+log ^{*}n)-round algorithm, where W is the largest node-weight (this bound translates to O(log² n⋅log (1/ε)) under the common assumption that W = poly(n)). As for α-dependent approximations, a deterministic CONGEST (8(1+ε)⋅α)-approximation algorithm with runtime O(log³ n⋅log (1/ε)) can be derived by combining the aforementioned algorithm of Faour et al. with a method presented by Kawarabayashi et al. [DISC 2020]. As our main results, we show the following.
- A deterministic CONGEST algorithm that computes an α^{1+τ}-approximation for MWIS in O(log nlog α) rounds for any constant τ > 0. To the best of our knowledge, this is the fastest runtime of any deterministic non-trivial approximation algorithm for MWIS to date. Furthermore, for the large class of graphs where α = Δ^{1-Θ(1)}, it implies a deterministic Δ^{1-Θ(1)}-approximation algorithm with a runtime of O(log nlog α) which improves upon the result of Faour et al. in both approximation ratio (by a Δ^{Θ(1)} factor) and runtime (by an O(log n/log α) factor).
- A deterministic CONGEST algorithm that computes an O(α)-approximation for MWIS in O(α^{τ}log n) rounds for any (desirably small) constant τ > 0. This improves the runtime of the best known deterministic O(α)-approximation algorithm in the case that α = O(polylog n). This also leads to a deterministic Δ(1+ε)-approximation algorithm with a runtime of O(α^{τ}log nlog (1/ε)) which improves upon the runtime of Faour et al. in the case that α = O(polylog n).
- A deterministic CONGEST algorithm that computes a (⌊(2+ε)α⌋)-approximation for MWIS in O(αlog n) rounds. This improves upon the best known α-dependent approximation ratio by a constant factor.
- A deterministic CONGEST algorithm that computes a 2d²-approximation for MWIS in time O(d²+log ^{*}n) in a directed graph with out-degree at most d. The dependency on n is (asymptotically) optimal due to a lower bound by Czygrinow et al. [DISC 2008] and Lenzen and Wattenhofer [DISC 2008]. We note that a key ingredient to all of our algorithms is a novel deterministic method that computes a high-weight subset of nodes whose induced subgraph is sparse.
https://drops.dagstuhl.de/storage/00lipics/lipics-vol286-opodis2023/LIPIcs.OPODIS.2023.16/LIPIcs.OPODIS.2023.16.pdf
Approximation algorithms
Sparse graphs
The CONGEST model