[Paper Review] Approximation algorithms for node-weighted prize-collecting Steiner tree problems on planar graphs

Jarosław Byrka, Mateusz Lewandowski+1 more|arXiv (Cornell University)|Jan 11, 2016

Complexity and Algorithms in Graphs2 references3 citations

TL;DR

This paper presents a new primal-dual Lagrangian-multiplier-preserving (LMP) 3-approximation algorithm for the node-weighted prize-collecting Steiner tree problem on planar graphs, followed by a (2.88 + ε)-approximation using threshold rounding and a 2.4-approximation for the node-weighted Steiner forest problem. The approach combines LMP analysis with randomized threshold rounding to improve the best-known approximation ratio without requiring LP solving.

ABSTRACT

We study the prize-collecting version of the Node-weighted Steiner Tree problem (NWPCST) restricted to planar graphs. We give a new primal-dual Lagrangian-multiplier-preserving (LMP) 3-approximation algorithm for planar NWPCST. We then show a ($2.88 + ε$)-approximation which establishes a new best approximation guarantee for planar NWPCST. This is done by combining our LMP algorithm with a threshold rounding technique and utilizing the 2.4-approximation of Berman and Yaroslavtsev for the version without penalties. We also give a primal-dual 4-approximation algorithm for the more general forest version using techniques introduced by Hajiaghay and Jain.

Motivation & Objective

• To develop a constant-factor approximation algorithm for the node-weighted prize-collecting Steiner tree problem (NWPCST) on planar graphs.
• To improve the best-known approximation ratio for NWPCST in planar graphs beyond previous bounds.
• To design a primal-dual algorithm that preserves Lagrangian multipliers for better integrality gap analysis.
• To combine the LMP algorithm with threshold rounding to achieve a tighter approximation guarantee.
• To extend the approach to the more general node-weighted prize-collecting Steiner forest (NWPCSF) problem on planar graphs.

Proposed method

• Adapt the primal-dual framework of Goemans and Williamson for edge-weighted PCST to the node-weighted case with a modified pruning phase.
• Separate analysis of connection and penalty costs by charging penalties to a dual solution part, ensuring Lagrangian-multiplier-preservation.
• Use threshold rounding with a random threshold α ∈ [0, β] to select terminals for a subproblem, then apply a 2.4-approximation for the resulting NWST instance.
• Combine the LMP 3-approximation with threshold rounding via a randomized selection between the two solutions to balance expected cost.
• Optimize the threshold β and mixing probability p to minimize the expected approximation ratio.
• Leverage the 2.4-approximation for planar node-weighted Steiner forest (NWSF) by Berman and Yaroslavtsev as a subroutine in the threshold rounding phase.

Experimental results

Research questions

• RQ1 Can a primal-dual LMP algorithm be designed for the node-weighted prize-collecting Steiner tree problem on planar graphs with improved approximation guarantees?
• RQ2 Can threshold rounding techniques be effectively combined with an LMP primal-dual algorithm to achieve a better approximation ratio than existing methods?
• RQ3 What is the best achievable approximation ratio for NWPCST on planar graphs using a combination of LMP and threshold rounding?
• RQ4 Can the approach be extended to the more general node-weighted prize-collecting Steiner forest problem on planar graphs?
• RQ5 How does the performance of the new algorithm compare to previous LP-based methods in terms of approximation ratio and computational requirements?

Key findings

• A new primal-dual 3-approximation algorithm is proposed for NWPCST on planar graphs, which is LMP and does not require solving an LP.
• By combining the LMP algorithm with threshold rounding and the 2.4-approximation for NWSF, a (2.88 + ε)-approximation is achieved for NWPCST on planar graphs.
• The approximation ratio of (2.88 + ε) improves upon the previous best-known bound of 2.93 + ε, which required solving an LP.
• A direct primal-dual 4-approximation algorithm is developed for the node-weighted prize-collecting Steiner forest problem on planar graphs.
• The analysis uses randomized rounding and expected cost bounds via integration over a random threshold α ∈ [0, β], with optimized parameters yielding the final ratio.
• The final approximation ratio is bounded by approximately 2.8797 + ε′, derived from the expression 4 / (4 - 3e^(-5/36)) × (1 + ε)², which is tighter than prior results.