Partial Connected Red-Blue Dominating Set

• Partial Connected Red-Blue Dominating Set is a framework that generalizes coverage problems by integrating connectivity constraints within a bipartite graph setting.
• It employs fixed-parameter tractable algorithms and parameterized approximation schemes to overcome inapproximability and W[1]-hardness challenges.
• The approach leverages Steiner out-tree reductions and twin-width based dynamic programming to provide efficient solutions under various graph restrictions.

The Partial Connected Red-Blue Dominating Set (ConnRBDS) problem is a unifying formalism for a range of coverage problems under connectivity constraints in graphs. It generalizes classical maximum coverage, partial dominating set, and partial vertex cover problems by introducing a bipartite framework with an auxiliary connectivity requirement, tightly connecting the domains of parameterized complexity, approximation algorithms, and structural graph theory (Inamdar et al., 13 Jan 2026).

1. Problem Definition and Formal Framework

ConnRBDS is defined over two graphs sharing a common “red” vertex set $R$ and a set $B$ of “blue” vertices:

• Incidence (Coverage) Graph: $G_{\rm cov} = (R \uplus B, E_{\rm cov})$, a bipartite graph where $(r, b) \in E_{\rm cov}$ if red vertex $r$ dominates blue vertex $b$.
• Connectivity (Constraint) Graph: $G_{\rm conn} = (R, E_{\rm conn})$, expressing structural requirements (e.g., connectivity) on red vertices.

Instance: $G_{\rm cov}$, $G_{\rm conn}$, integers $k$ and $t$.

Question: Does there exist $S \subseteq R$ such that $|S| \le k$, $G_{\rm conn}[S]$ is connected, and $|N_{G_{\rm cov}}(S)| \ge t$?

Two canonical optimization variants are frequently studied:

• Maximum Coverage under Connectivity: $$\max \{ |N_{G_{\rm cov}}(S)| : S \subseteq R, |S| \le k, G_{\rm conn}[S] \text{ is connected} \}$$.
• Minimum Size for Target Coverage: $$\min \{ |S| : S \subseteq R, |N_{G_{\rm cov}}(S)| \ge t, G_{\rm conn}[S] \text{ is connected} \}$$.

An $(\alpha, \beta)$-approximation yields $S$ with $|S| \le \alpha k$ and $|N_{G_{\rm cov}}(S)| \ge \beta t$, or reports infeasibility if the exact $(1,1)$-goal is impossible.

2. Hardness Landscape

ConnRBDS inherits significant inapproximability and parameterized hardness from classical coverage problems:

• Classical Inapproximability: No polynomial-time $(1, 1-\tfrac{1}{e}+\varepsilon)$-approximation exists for coverage, unless $P=NP$. This barrier propagates to ConnRBDS via reductions from the classical Max Coverage problem (Inamdar et al., 13 Jan 2026).
• $\W[1]$-Hardness Parameterized by $k$: Even when $G_{\rm conn}$ is a clique (no real connectivity constraint) or star, and $G_{\rm cov}$ is 3-degenerate, ConnRBDS is $\W[1]$-hard in $k$. This is shown by a reduction from Partial Dominating Set on 2-degenerate graphs.
• ETH-Based Lower Bounds: No $f(k)\,n^{o(k)}$-time $(g(k), 1)$-approximation is achievable (size-side) unless ETH fails. Reductions from the Dominating Set problem enforce this barrier.
• ConnPVC Variant: Restricting blue vertices to degree 2 (ConnPVC) remains $\W[1]$-hard [Guo–Niedermeier–Wahlström WG 2007].

A summary of the hardness results is presented below:

Problem Variant	Hardness Class	Restriction
Max Coverage under Connectivity	$(1, 1-\tfrac{1}{e}+\varepsilon)$-inapprox., unless $P=NP$	General graphs
Parameterized by $k$	$\W[1]$-hard	$G_{\rm conn}$ clique/star, $G_{\rm cov}$ 3-degenerate
ConnPVC (degree 2 blues)	$\W[1]$-hard	All blue degrees 2

3. Fixed-Parameter Tractability by Coverage Target $t$

Despite the $\W[1]$-hardness in $k$, ConnRBDS is fixed-parameter tractable (FPT) when parameterized by the coverage threshold $t$.

• Steiner Out-Tree Reduction: The problem reduces in FPT time to the (Relaxed) Directed Steiner Out-Tree problem, which seeks an out-tree covering a set of terminals within vertex/edge budget. This intermediate formulation allows the use of an FPT algorithm running in $O^*((2e)^t)$ randomized (or $O^*((2e)^{t+o(t)})$ deterministic) time (Inamdar et al., 13 Jan 2026).
• Correctness: There is a correspondence between solutions of ConnRBDS and Steiner out-trees spanning terminals associated with covered blues, respecting connectivity and cardinality constraints.

This suggests that, even with arbitrary connectivity constraints, progress is possible in the parameter $t$.

4. Parameterized Approximation Schemes for Biclique-Free Incidence Graphs

For incidence graphs $G_{\rm cov}$ excluding $K_{d,d}$ (biclique-free), ConnRBDS admits two complementary parameterized approximation frameworks.

4.1 Efficient Parameterized Approximation Scheme (EPAS) for Coverage

• Input: $(G_{\rm cov}, G_{\rm conn}, k, t)$, $G_{\rm cov}$ $K_{d,d}$-free, $\varepsilon > 0$.
• Guarantee: In $$2^{O\!\left(\frac{k^2 d}{\varepsilon}\right)} n^{O(1)}$$ time, either return a connected $S \subseteq R$, $|S| \le k$, $|N_{G_{\rm cov}}(S)| \ge (1-\varepsilon)t$, or assert impossibility for $t$.
• Key Steps: Construction of a conflict graph on $R$, random separation via lopsided-universal families, component filtering, edge sparsification within components, weight assignment on red vertices, and max-weight $k$-tree subtree enumeration in $G_{\rm conn}$ (with established enumeration and dynamic programming schemes).

4.2 Parameterized Approximation Scheme (PAS) for Size Minimization

• Input: $(G_{\rm cov}, G_{\rm conn}, k, t)$, $G_{\rm cov}$ $K_{d,d}$-free, $\varepsilon > 0$.
• Guarantee: In $$2^{O\left(k d(k^2+\log d)\right)} n^{O(1/\varepsilon)}$$ time, obtain a connected $S \subseteq R$, $|S| \le (1+\varepsilon)k$, covering at least $t$ blue vertices.
• Techniques: Seed-cover lemma to identify small hub sets, recursive branching on seed selection, localized EPAS-instances, and path augmentation for high-degree case analysis.

A table summarizing these schemes:

Scheme	Objective	Time Complexity	Guarantee
EPAS	Approx. max coverage	$$2^{O\left(\frac{k^2 d}{\varepsilon}\right)} n^{O(1)}$$	$(k, (1-\varepsilon)t)$
PAS	Approx. min size	$2^{O(k d (k^2 + \log d))} n^{O(1/\varepsilon)}$	$((1+\varepsilon)k, t)$

Crucially, these schemes do not restrict $G_{\rm conn}$, generalizing prior coverage approximations to the connectivity-constrained setting (Inamdar et al., 13 Jan 2026).

5. FPT Algorithms on Bounded Twin-Width Graphs

ConnRBDS is also fixed-parameter tractable in $(\text{twin-width}(G), k)$ for graphs with bounded twin-width given a contraction sequence:

• Formalization: The property "exists $X \subseteq R$, $|X|=k$, $G[X]$ connected, dominating $\geq t$ blue vertices" is expressible in first-order logic with counting quantifiers (FO+counting).
• Algorithmic Strategy: Executes dynamic programming over the contraction sequence of width $d$. Maintains extended profile tables, propagates connectivity and coverage, and ensures that at each step profiles have at most $k(d+1)$ clusters, yielding $2^{O(kd \log(kd))}$ table size at each level.
• Complexity: $O(2^{O(kd \log(kd))}\,n)$ time, thus FPT in $(d, k)$ (Balabán et al., 25 Apr 2025).

A plausible implication is that a broad class of structural graph constraints can be accommodated using this logical and DP framework.

6. Connections, Extensions, and Open Directions

ConnRBDS serves as a unifying model, clarifying the boundary between inapproximability and FPT-approximability for coverage problems under connectivity constraints (Inamdar et al., 13 Jan 2026).

• Neighborhood Sparsifiers: The use of sparsifiers—small families of weight functions—facilitates reduction of coverage objectives to weighted connectivity, a tool with potential in broader scenarios.
• Dichotomy: For unrestricted $G_{\rm cov}$, FPT approximations better than $(1, 1-1/e)$ for coverage are precluded under ETH. For $K_{d,d}$-free incidences, parameterized schemes become available.
• Other Constraints: Generalization to independence, matroid, vertex-connectivity, or edge-capacity constraints in $G_{\rm conn}$ is possible via the same two-graph paradigm.
• Compression and Kernels: ConnPVC’s $\W[1]$-hardness obviates $\alpha$-approximate polynomial compression, contrasting with unconstrained Max Coverage where approximation kernels are possible.

This suggests that the analytic methodologies and combinatorial principles of ConnRBDS will continue to inform algorithmic advances in both theoretical and applied network design, sensor placement, and coverage planning in large-scale graphs.