CompGS Framework: Graph Complexity Dichotomy

• CompGS Framework is a meta-classification system that evaluates the computational complexity of graph problems on ℋ-subgraph-free graphs using forbidden subgraph structures.
• It employs a dichotomy theorem showing that the presence of subdivided claws or paths guarantees bounded treewidth and polynomial-time solvability, while their absence implies NP-hardness.
• The framework unifies classical complexity results for various graph problems and guides future exploration in fine-grained algorithmic and hardness analyses.

The CompGS framework, in the context of computational complexity and forbidden subgraphs, is a meta-theoretic classification system that provides a unified methodology for determining the computational complexity of a large class of graph problems when restricted to subgraph-free graph classes defined by a finite set ℋ of forbidden subgraphs. The framework, proposed by Johnson et al. in "Complexity Framework For Forbidden Subgraphs I: The Framework", formalizes conditions under which the efficiency of solving graph problems transitions sharply between polynomial-time tractability and NP-hardness, based solely on the structural implications of the forbidden subgraph set ℋ (Johnson et al., 2022).

1. Formal Definitions and Core Concepts

Let ℋ denote a finite set of graphs. A graph $G$ is termed ℋ-subgraph-free if no $H \in ℋ$ is a (not necessarily induced) subgraph of $G$, i.e., $H$ cannot be obtained from $G$ via vertex and/or edge deletions.

A key structural component is the set $\mathcal{S}$, comprising all nonempty disjoint unions of subdivided claws and paths:

• A path $P_k$ is a simple $k$-vertex path.
• A subdivided claw $S_{p,x,r}$ is constructed from the 3-leaf star $K_{1,3}$ by subdividing its three edges $H \in ℋ$0, $H \in ℋ$1, and $H \in ℋ$2 times respectively.
• Every graph in $H \in ℋ$3 has max degree at most 3; the connected graphs in $H \in ℋ$4 are exactly the paths and subdivided claws $H \in ℋ$5.

A problem $H \in ℋ$6 is classified as a C123-problem if it meets all the following:

• C1: Polynomial-time solvable on all (not necessarily hereditary) graph classes of bounded treewidth (often even in linear time).
• C2: NP-hard even when restricted to subcubic graphs (graphs of maximum degree 3).
• C3: NP-hardness is preserved under edge subdivisions; there exists $H \in ℋ$7 such that, if $H \in ℋ$8 is NP-hard on a class of subcubic graphs $H \in ℋ$9, then for every $G$0 it remains NP-hard on the class $G$1, produced by $G$2-subdividing every edge of each graph in $G$3.

2. Meta-classification Theorem and Dichotomy

The CompGS meta-classification theorem states: For any C123-problem $G$4 and any finite forbidden set $G$5, the algorithmic complexity of $G$6 on $G$7-subgraph-free graphs is determined by whether $G$8 "forces" bounded treewidth—that is, whether $G$9.

• If $H$0 (i.e., $H$1 contains a path or subdivided claw, so the $H$2-subgraph-free graph class has bounded treewidth), then $H$3 is polynomial-time solvable on $H$4-subgraph-free graphs.
• If $H$5 (so $H$6 does not promote bounded treewidth), then $H$7 is NP-hard on $H$8-subgraph-free graphs.

The theorem leverages classical graph minor theory: $H$9 if and only if the class of $G$0-subgraph-free graphs has bounded pathwidth (and thus bounded treewidth), per the Robertson–Seymour and Bienstock–Robertson–Seymour–Thomas results.

Fine-grained variants also hold: for some problems, an almost-linear vs. no subquadratic algorithm dichotomy can be exhibited under computational hardness conjectures.

3. Proof Methodology and Structural Rationale

The proof of the main dichotomy theorem operates in two regimes:

Algorithmic Side ($G$1):

If $G$2 contains a path or subdivided claw, every $G$3-subgraph-free $G$4 is $G$5-minor-free for some $G$6. The Robertson–Seymour theory ensures bounded treewidth, so the C1 condition guarantees polynomial-time algorithms for $G$7.

Hardness Side ($G$8):

In this case, each $G$9 contains structures (cycles, $\mathcal{S}$0, subdivided "H-graphs") ensuring unbounded treewidth. Sophisticated gadget reductions, building on techniques of Alekseev–Korobitsyn, Kamiński, and Golovach–Paulusma, show that any C123-problem remains NP-hard even if small cycles, $\mathcal{S}$1, and the subdivided H-graphs are forbidden for sufficiently large parameters, so long as $\mathcal{S}$2 does not contain any of the graphs in $\mathcal{S}$3.

4. Application to Classical Graph Problems

Several widely studied problems fall under the CompGS C123 regime, enjoying the dichotomy:

Category	Problems	C1: DP on bounded-tw	C2: subcubic NP-hard	C3: subdivision preserves hardness
Width-parameters	Treewidth, Pathwidth	✓	✓	✓
Packing/Covering	Independent Set, Vertex Cover, Dominating Set, Edge Dominating Set, Odd Cycle Transversal, List Colouring, P₃-Factor	✓	✓	✓
Network Design	Max-Cut, (Edge/Node) Steiner Tree, Multiway Cut, Disjoint Paths, Long/Induced Path	✓	✓	✓
Fine-grained	Diameter, Radius (under OV/HS)	✓	✓	✓

For these, the CompGS framework enables instantaneous determination of polynomial-time solvability or NP-hardness on any $\mathcal{S}$4-subgraph-free class once the intersection with $\mathcal{S}$5 is checked. Some problems also inherit sharp fine-grained lower bounds (e.g., no $\mathcal{S}$6 algorithms for Diameter/Radius unless OV or Hitting Set fails).

5. Open Problems and Directions for Extension

Several lines of inquiry remain for extending or refining the CompGS framework:

• Infinite $\mathcal{S}$7: While the meta-classification is established for finite $\mathcal{S}$8, a generalization to certain natural infinite $\mathcal{S}$9 remains open; the compactness arguments of minor theory do not generally carry over.
• Relaxing Condition C3: The core proofs demand hardness persistence under some $P_k$0-subdivision in subcubic graphs. Exploring relaxations or alternative forms of C3 compatible with a dichotomy theorem is an ongoing topic.
• Natural Problems Satisfying Only C1 and a Weaker C3: Instances exist where problems have polynomial-time algorithms on bounded treewidth and NP-hardness persists under subdivision but not on all subcubic graphs; the existence of natural examples remains unclear.
• Induced-Subgraph Version: A richer structural theory exists for hereditary (induced-subgraph-free) graph classes. Recent work (Lozin–Razgon) characterizes bounded treewidth in this domain, suggesting potential for a corresponding dichotomy.
• Partial Condition Failures: If a problem fails one of C1, C2, or C3 (e.g., Hamiltonian Cycle is easy under subdivision, violating C3), partial subclassifications are possible. Subsequent works (III, IV, …) analyze such intermediate regimes.

6. Significance and Impact

The CompGS framework offers a “single–theorem pathway” to resolve the complexity landscape for a substantial swath of classical graph problems over arbitrary finite forbidden-subgraph classes. By reducing the complexity classification to a purely structural check (does $P_k$1 meet $P_k$2?), it provides an algorithmic and theoretical tool to unify and generalize a long sequence of bespoke hardness and tractability results scattered across the literature (Johnson et al., 2022).

This perspective highlights treewidth as a central graph invariant governing computational phase transitions in a wide range of problems and formalizes the empirical phenomena that were previously verified only on a case-by-case basis. The framework resolves longstanding open questions about when such problems become efficiently solvable versus universally hard, and motivates further exploration into the richness of forbidden subgraph theory, particularly regarding hereditary graph classes and finer algorithmic distinctions.