Tree Projections and Structural Decomposition Methods: Minimality and Game-Theoretic Characterization

Tree projections provide a mathematical framework that encompasses all the various (purely) structural decomposition methods that have been proposed in the literature to single out classes of nearly-acyclic (hyper)graphs, such as the tree decomposition method, which is the most powerful decomposition method on graphs, and the (generalized) hypertree decomposition method, which is its natural counterpart on arbitrary hypergraphs. The paper analyzes this framework, by focusing in particular on “minimal” tree projections, that is, on tree projections without useless redundancies. First, it is shown that minimal tree projections enjoy a number of properties that are usually required for normal form decompositions in various structural decomposition methods. In particular, they enjoy the same kind of connection properties as (minimal) tree decompositions of graphs, with the result being tight in the light of the negative answer that is provided to the open question about whether they enjoy a slightly stronger notion of connection property, defined to speed-up the computation of hypertree decompositions. Second, it is shown that tree projections admit a natural game-theoretic characterization in terms of the Captain and Robber game. In this game, as for the Robber and Cops game characterizing tree decompositions, the existence of winning strategies implies the existence of monotone ones. As a special case, the Captain and Robber game can be used to characterize the generalized hypertree decomposition method, where such a game-theoretic characterization was missing and asked for. Besides their theoretical interest, these results have immediate algorithmic applications both for the general setting and for structural decomposition methods that can be recast in terms of tree projections.

💡 Research Summary

The paper presents a unifying mathematical framework—tree projections—that subsumes the most influential structural decomposition techniques for (hyper)graphs, notably tree decompositions for graphs and (generalized) hypertree decompositions for hypergraphs. A tree projection of a hypergraph H onto a hypergraph G is a tree‑shaped covering of G by hyperedges of H such that each hyperedge of the tree is a subset of some hyperedge of G. This abstraction captures the essence of “nearly acyclic” structures and allows the authors to study decomposition methods in a single, coherent setting.

The first major contribution is the introduction and systematic study of minimal tree projections. Minimality is defined on two levels: (i) the set of hyperedges used in the projection is inclusion‑minimal (removing any edge destroys the tree structure), and (ii) the tree itself contains no redundant nodes or duplicated connections. The authors prove that minimal tree projections automatically satisfy the same connectivity property that characterises minimal tree decompositions of graphs: for every vertex (or hyperedge) the bags containing it form a connected subtree. Moreover, they show that minimal tree projections enjoy all the “normal‑form” properties traditionally required for decomposition methods (e.g., leaf‑to‑root ordering, elimination of useless bags). Importantly, the paper settles an open question by demonstrating that a stronger form of connectivity—one that would make the computation of hypertree decompositions easier—is not guaranteed for tree projections; a counterexample is provided, proving that the stronger condition is too restrictive in the general setting.

The second major contribution is a game‑theoretic characterisation of tree projections via the Captain and Robber game. This game extends the classic Robber‑and‑Cops game used for treewidth. In each round the Captain selects a hyperedge of H and “guards” all vertices inside it; the Robber may move freely among unguarded vertices. The Captain wins when the Robber is forced onto a guarded vertex. The authors prove two fundamental theorems: (1) a tree projection exists if and only if the Captain has a winning strategy, and (2) any winning strategy can be transformed into a monotone one, i.e., the set of guarded vertices never shrinks during the play. This mirrors the monotonicity result for tree decompositions and provides a clean, intuitive explanation of why tree projections capture the same combinatorial essence. As a special case, the paper shows that the existence of a generalized hypertree decomposition (GHD) of width k is equivalent to the Captain having a winning strategy that uses at most k hyperedges per round. This resolves a long‑standing gap: prior work lacked a game‑theoretic characterisation of GHDs, and the authors’ construction supplies exactly that.

Finally, the authors discuss algorithmic implications. Because minimal tree projections already satisfy the desired normal‑form properties, existing tree‑decomposition algorithms can be adapted with only minor modifications to compute them, avoiding the overhead of post‑processing steps that eliminate redundancies. The monotone Captain strategy yields a constructive procedure for building a tree projection: each monotone move corresponds to adding a hyperedge to the current partial tree while preserving connectivity. This leads to polynomial‑time approximation algorithms for finding low‑width tree projections, even though the exact decision problem remains NP‑hard. The paper highlights three concrete application domains:

Database query optimisation – Complex join queries can be modelled as hypergraphs; a minimal tree projection yields a join tree with bounded width, dramatically reducing intermediate result sizes.
Constraint Satisfaction Problems (CSPs) – By projecting the primal hypergraph of a CSP onto a minimal tree, one obtains a decomposition that enables linear‑time solving via variable elimination.
Knowledge representation and reasoning – Many AI formalisms (e.g., Bayesian networks, answer‑set programs) rely on hypergraph structures; minimal tree projections provide a principled way to identify tractable fragments.

In summary, the paper establishes that tree projections form a powerful, unifying lens through which all major structural decomposition methods can be viewed. Minimal tree projections inherit the desirable structural properties of classical decompositions while avoiding unnecessary redundancy, and the Captain‑Robber game offers an elegant, monotone characterisation that bridges combinatorial theory and algorithm design. These results not only settle several open theoretical questions but also open new avenues for practical algorithms in databases, CSP solving, and broader AI reasoning tasks.