[Paper Review] An approximation algorithm for the Bandpass-2 problem

Weitian Tong, Zhizhong Chen+5 more|arXiv (Cornell University)|Jul 26, 2013

graph theory and CDMA systems12 references3 citations

TL;DR

This paper presents a novel $\frac{426}{227}$-approximation algorithm for the Bandpass problem, an NP-hard optimization problem in optical communication networks that seeks to maximize bandpass extraction via row permutation of a binary matrix. The algorithm leverages maximum weight $b$-matchings for $b=1,2,4$, exploits an intrinsic structural property of optimal solutions, and introduces advanced partitioning schemes to decompose 2-matchings and 4-matchings into acyclic components, achieving the best-known performance ratio to date.

ABSTRACT

The general Bandpass-$B$ problem is NP-hard and can be approximated by a reduction into the weighted $B$-set packing problem, with a worst case performance ratio of $O(B^2)$. When $B = 2$, a maximum weight matching gives a 2-approximation to the problem. In this paper, we call the Bandpass-2 problem simply the Bandpass problem. The Bandpass problem can be viewed as a variation of the maximum traveling salesman problem, in which the edge weights are dynamic rather than given at the front. We present a ${426}{227}$-approximation algorithm for the problem. Such an improved approximation is built on an intrinsic structural property proven for the optimal solution and several novel schemes to partition a $b$-matching into desired matchings.

Motivation & Objective

• To develop a better approximation algorithm for the Bandpass problem, which is NP-hard and arises in optimizing wavelength division multiplexing in optical networks.
• To improve upon the previous best-known approximation ratio of $\frac{220}{117} \approx 1.8805$ for the Bandpass-2 problem.
• To establish a tighter worst-case performance ratio by leveraging structural insights into optimal row permutations and novel partitioning techniques for $b$-matchings.
• To design an algorithm with $O(n^4)$ time complexity that guarantees a performance ratio of $\frac{426}{227} \approx 1.8767$.

Proposed method

• The algorithm constructs an edge-weighted graph from the input binary matrix and computes a maximum weight matching $M_1$ for $b=1$.
• It computes a second maximum weight matching $M_2$ in the residual graph, ensuring the union of $M_1$ and $M_2$ is acyclic via a novel 2-matching partitioning scheme.
• A third maximum weight $4$-matching is computed and partitioned into 7.5 desired matchings to maintain acyclicity when combined with $M_1$.
• The method relies on a structural property of the optimal solution, showing that certain bandpasses in the optimum are tightly linked to the first matching $M_1$.
• Lemmas are used to bound the weight of $M_2$ in terms of $|B_4|$ (bandpasses of length 4), proving $w'(M_2) \geq \frac{7}{27}|B_4|$.
• A linear combination of bounds from $M_1$, $M_2$, and structural components is used to derive the final performance ratio via optimization over parameters $x$ and $y$.

Experimental results

Research questions

• RQ1 Can a tighter approximation ratio be achieved for the Bandpass problem by exploiting deeper structural properties of optimal row permutations?
• RQ2 Can advanced partitioning schemes for $b$-matchings ($b=2,4$) be designed to preserve acyclicity and improve approximation guarantees?
• RQ3 Is it possible to improve upon the $\frac{220}{117}$-approximation by combining multiple maximum weight matchings with refined analysis?
• RQ4 How does the dynamic nature of edge weights in the Bandpass problem affect the potential for achieving ratios close to those of the Max-TSP?

Key findings

• The proposed algorithm achieves a worst-case performance ratio of $\frac{426}{227} \approx 1.8767$, improving upon the prior best of $\frac{220}{117} \approx 1.8805$.
• The algorithm runs in $O(n^4)$ time, dominated by the computation of maximum weight $b$-matchings for $b=1,2,4$.
• A key structural property is proven: the optimal solution contains bandpasses that are tightly linked to the first maximum weight matching $M_1$.
• The weight of the second matching $M_2$ is shown to be at least $\frac{7}{27}$ of the number of length-4 bandpasses in the optimal solution.
• The analysis combines bounds from $M_1$, $M_2$, and structural components using optimized parameters $x = \frac{35}{71}$ and $y = \frac{57}{71}$ to derive the final ratio.
• The result demonstrates that $b$-matching partitioning and acyclicity enforcement are critical for improving approximation ratios in dynamic-weight problems like the Bandpass problem.