An Effective Dichotomy for the Counting Constraint Satisfaction Problem

Bulatov (2008) gave a dichotomy for the counting constraint satisfaction problem #CSP. A problem from #CSP is characterised by a constraint language, which is a fixed, finite set of relations over a finite domain D. An instance of the problem uses these relations to constrain the variables in a larger set. Bulatov showed that the problem of counting the satisfying assignments of instances of any problem from #CSP is either in polynomial time (FP) or is #P-complete. His proof draws heavily on techniques from universal algebra and cannot be understood without a secure grasp of that field. We give an elementary proof of Bulatov’s dichotomy, based on succinct representations, which we call frames, of a class of highly structured relations, which we call strongly rectangular. We show that these are precisely the relations which are invariant under a Mal’tsev polymorphism. En route, we give a simplification of a decision algorithm for strongly rectangular constraint languages, due to Bulatov and Dalmau (2006). We establish a new criterion for the #CSP dichotomy, which we call strong balance, and we prove that this property is decidable. In fact, we establish membership in NP. Thus, we show that the dichotomy is effective, resolving the most important open question concerning the #CSP dichotomy.

💡 Research Summary

The paper revisits the celebrated dichotomy theorem for the counting constraint satisfaction problem (#CSP) originally proved by Bulatov (2008). While Bulatov’s proof relies heavily on universal algebra—particularly on the structure of congruence lattices and the notion of “congruence singularity”—the present work offers an elementary, self‑contained alternative that avoids deep algebraic machinery and, crucially, shows that the dichotomy is effective, i.e., decidable in NP.

The authors introduce the class of strongly rectangular relations, which are exactly those relations preserved by a Mal’tsev polymorphism. For such relations they define a compact representation called a frame. A frame is a subset F of a relation R⊆Dⁿ satisfying two conditions: (1) every tuple of R appears in F, and (2) for each coordinate i, the collection of i‑equivalent sets (sets of possible values given a common prefix) is the same in R and in F. Frames are dramatically smaller than the full relation (size at most |D|ⁿ) yet retain enough structure to reconstruct R completely. The paper provides a polynomial‑time algorithm to construct a frame for the solution set of any instance whose constraint language Γ is strongly rectangular, and proves that the frame is empty iff the instance has no satisfying assignment. This yields a concrete O(m·n⁴) decision algorithm for CSP(Γ) (m constraints, n variables), improving on the earlier Bulatov‑Dalmau result which gave only a polynomial‑time existence claim.

To obtain a counting algorithm, the authors define a new criterion called strong balance. A ternary relation R⊆A₁×A₂×A₃ is balanced if, after permuting rows and columns, its associated matrix becomes block‑diagonal with each block of rank one. An n‑ary relation is balanced if every way of viewing it as a ternary relation (by partitioning its coordinates into three groups) is balanced. A constraint language Γ is strongly balanced if every pp‑definable relation of arity ≥3 is balanced. The paper proves that strong balance is equivalent to Bulatov’s congruence singularity condition.

If Γ is strongly balanced, the counting algorithm proceeds as follows. From the frame of the solution set R, the algorithm computes, for each ordered pair of indices (i, j) and each domain element a, the number N_{i,j}(a) of partial assignments where the i‑th prefix is fixed and the j‑th variable takes value a. These numbers satisfy simple recurrence relations that can be evaluated using the rank‑one block structure of the matrices derived from the balanced ternary projections. By iteratively computing N_{i,j} for increasing i, the algorithm finally obtains N_{n‑1,n}(a) for all a, and sums them to get |R|, the total number of satisfying assignments. The crucial observation is that for a rank‑one block matrix, the row and column sums uniquely determine the entire matrix, allowing reconstruction without enumerating exponentially many entries. Consequently, for strongly balanced languages #CSP(Γ) is solvable in polynomial time; otherwise, the problem is #P‑complete.

The final contribution is a decidability proof for strong balance. The authors show that if a language Γ fails the strong balance condition, there exists a counterexample instance whose size is polynomial in |Γ|. By reformulating the balance condition as a counting problem on a derived formula and applying a technique of Lovász, they reduce the decision to checking the existence of a certain symmetry in a fixed structure computable from Γ. This reduction yields an NP algorithm for deciding strong balance, establishing that the dichotomy is effective.

In summary, the paper delivers:

1. An elementary, algebra‑free proof of Bulatov’s #CSP dichotomy.
2. The notion of frames for strongly rectangular relations, together with explicit construction and reconstruction algorithms.
3. A new, equivalent criterion—strong balance—under which counting becomes tractable.
4. A polynomial‑time counting algorithm for the tractable cases, matching the decision‑time complexity.
5. An NP‑membership proof for the meta‑problem of determining which side of the dichotomy a given finite constraint language falls on.

These results close the major open question about the effectiveness of the #CSP dichotomy and provide tools that are both theoretically elegant and practically implementable.