A $5$-Approximation Analysis for the Cover Small Cuts Problem

In the Cover Small Cuts problem, we are given a capacitated (undirected) graph $G=(V,E,u)$ and a threshold value $λ$, as well as a set of links $L$ with end-nodes in $V$ and a non-negative cost for each link $\ell\in L$; the goal is to find a minimum-cost set of links such that each non-trivial cut of capacity less than $λ$ is covered by a link. Bansal, Cheriyan, Grout, and Ibrahimpur (arXiv:2209.11209, Algorithmica 2024) showed that the WGMV primal-dual algorithm, due to Williamson, Goemans, Mihail, and Vazirani (Combinatorica, 1995), achieves approximation ratio $16$ for the Cover Small Cuts problem; their analysis uses the notion of a pliable family of sets that satisfies a combinatorial property. Later, Bansal (arXiv:2308.15714v2, IPCO 2025) and then Nutov (arXiv:2504.03910, MFCS 2025) proved that the same algorithm achieves approximation ratio $6$. We show that the same algorithm achieves approximation ratio $5$, by using a stronger notion, namely, a pliable family of sets that satisfies symmetry and structural submodularity.

💡 Research Summary

The paper revisits the “Cover Small Cuts” problem, where one is given an undirected capacitated graph G = (V,E,u), a threshold λ, and a set L of potential links each with a non‑negative cost. The objective is to select a minimum‑cost subset of links so that every non‑trivial cut whose total capacity is smaller than λ is intersected by at least one chosen link.

Earlier work showed that the primal‑dual algorithm of Williamson, Goemans, Mihail, and Vazirani (WGMV, 1995) yields a 16‑approximation (Bansal et al., 2022) and later a 6‑approximation (Bansal, 2023‑2025; Nutov, 2025) for this problem. Both analyses rely on the notion of a pliable family of sets: for any two sets A, B in the family, at least two of the four derived sets A∩B, A∪B, A\B, B\A also belong to the family.

The authors improve the approximation factor to 5 by strengthening the structural requirements on the pliable family. They require two additional properties:

1. Symmetry – if S belongs to the family then its complement V \ S also belongs.
2. Structural submodularity – for any crossing pair A, B, at least one of A∩B or A∪B lies in the family, and at least one of A\B or B\A lies in the family.

These conditions are exactly the “structural submodularity” introduced by Diestel et al. in the study of abstract separation systems.

The paper first proves that a family satisfying symmetry and structural submodularity is automatically pliable and enjoys several useful combinatorial properties:

• Lemma 1 shows that inclusion‑minimal sets are pairwise disjoint.
• Lemma 2 establishes the “sparse crossing” property: any set in the family crosses at most one inclusion‑minimal set.
• Lemma 3 and Proposition 6 extend the earlier property (γ) to a stronger version (γ★), which handles multiple disjoint subsets simultaneously.

With these tools the authors turn to crossing density ρ(F), a parameter introduced by Bansal. For a family F, ρ(F) is the smallest integer ρ such that for any edge set E′ and any laminar subfamily of F restricted to E′, the number of crossing pairs (S, C) (where S is a laminar set and C is an inclusion‑minimal uncovered set) is at most ρ·|C|. Bansal proved that the WGMV algorithm achieves an approximation ratio of 3 + ρ(F).

The core technical contribution is the proof that for any pliable family satisfying symmetry and structural submodularity, ρ(F) ≤ 2. The argument proceeds as follows:

• For a minimal covering set J of the uncovered inclusion‑minimal sets C_bE, the authors consider a laminar family L of witness sets (each witness set is a set in F covered by exactly one link of J).
• By the sparse crossing property, each witness set crosses at most one set in C_bE.
• They construct a rooted tree T whose nodes correspond to the witness sets together with the whole ground set V. Each node’s parent is the smallest witness set that strictly contains it.
• A mapping ψ assigns each C∈C_bE to the smallest node in T that contains C. Nodes that receive a mapping are colored red.
• Using Lemma 7, Lemma 8, and the strengthened property (γ★), they show that every non‑root node in T is either red itself or has a red child. Consequently, the number of red nodes is at most |C_bE|, and every red node accounts for at most two witness nodes (itself and possibly its parent). Hence the total number of witness nodes |L| ≤ 2·|C_bE|, which directly yields ρ(F) ≤ 2.

Since Nutov (2025) constructed an instance where the WGMV algorithm returns a solution of cost (5 − ε)·OPT for arbitrarily small ε > 0, the bound 5 is tight.

Thus the paper concludes that the classic WGMV primal‑dual algorithm, when applied to the Cover Small Cuts problem, attains a tight 5‑approximation. The analysis not only improves the known bound but also simplifies the combinatorial reasoning by leveraging symmetry, structural submodularity, and the crossing‑density framework. This result settles the long‑standing question of the exact approximation ratio for this algorithmic paradigm on the Cover Small Cuts problem.