Unfounded Sets and Well-Founded Semantics of Answer Set Programs with Aggregates
Logic programs with aggregates (LPA) are one of the major linguistic extensions to Logic Programming (LP). In this work, we propose a generalization of the notions of unfounded set and well-founded semantics for programs with monotone and antimonotone aggregates (LPAma programs). In particular, we present a new notion of unfounded set for LPAma programs, which is a sound generalization of the original definition for standard (aggregate-free) LP. On this basis, we define a well-founded operator for LPAma programs, the fixpoint of which is called well-founded model (or well-founded semantics) for LPAma programs. The most important properties of unfounded sets and the well-founded semantics for standard LP are retained by this generalization, notably existence and uniqueness of the well-founded model, together with a strong relationship to the answer set semantics for LPAma programs. We show that one of the D-well-founded semantics, defined by Pelov, Denecker, and Bruynooghe for a broader class of aggregates using approximating operators, coincides with the well-founded model as defined in this work on LPAma programs. We also discuss some complexity issues, most importantly we give a formal proof of tractable computation of the well-founded model for LPA programs. Moreover, we prove that for general LPA programs, which may contain aggregates that are neither monotone nor antimonotone, deciding satisfaction of aggregate expressions with respect to partial interpretations is coNP-complete. As a consequence, a well-founded semantics for general LPA programs that allows for tractable computation is unlikely to exist, which justifies the restriction on LPAma programs. Finally, we present a prototype system extending DLV, which supports the well-founded semantics for LPAma programs, at the time of writing the only implemented system that does so. Experiments with this prototype show significant computational advantages of aggregate constructs over equivalent aggregate-free encodings.
💡 Research Summary
The paper addresses a long‑standing gap in the semantics of logic programs that incorporate aggregate functions. While aggregates are a powerful extension of traditional logic programming (LP), the classic notions of unfounded sets and well‑founded semantics have been defined only for aggregate‑free programs. The authors close this gap by introducing a generalized definition of unfounded sets that works for programs containing monotone and antimonotone aggregates (referred to as LPAma programs). Their definition preserves the essential intuition of the original concept: an atom belongs to an unfounded set if, under the current partial interpretation, there is no rule that can justify it positively, and assuming its falsity does not make any rule’s body true. Importantly, when aggregates are absent, the definition collapses to the standard LP unfounded set, guaranteeing backward compatibility.
Building on this new unfounded‑set notion, the authors define a well‑founded operator (WF‑operator). The operator iteratively performs two steps on a partial interpretation I: (1) a positive expansion that adds atoms whose bodies are satisfied in I, and (2) a negative reduction that removes atoms that belong to a newly computed unfounded set with respect to I. Repeating these steps until a fixpoint is reached yields the well‑founded model of the program. The paper proves three fundamental properties of this construction: (i) existence and uniqueness of the well‑founded model for every LPAma program, (ii) a strong relationship with answer‑set semantics—every answer set is a superset of the well‑founded model, and if answer sets exist they contain the well‑founded model as a subset, and (iii) equivalence with the D‑well‑founded semantics introduced by Pelov, Denecker, and Bruynooghe. The latter result shows that two seemingly different formal frameworks (approximation operators versus unfounded sets) converge on the same semantics for the class of monotone/antimonotone aggregates.
From a computational standpoint, the authors demonstrate that the well‑founded model of an LPAma program can be computed in polynomial time. The key observation is that for monotone and antimonotone aggregates, checking the satisfaction of an aggregate atom with respect to a partial interpretation can be done efficiently, allowing the unfounded‑set computation to remain tractable. In contrast, they prove that for general LPA programs that may contain aggregates which are neither monotone nor antimonotone, the problem of deciding aggregate satisfaction under partial interpretations is coNP‑complete. Consequently, a tractable well‑founded semantics for the unrestricted class is unlikely, justifying the focus on LPAma programs.
To validate their theoretical contributions, the authors implemented a prototype extending the DLV system. The prototype supports the new well‑founded semantics, parses aggregate‑rich programs, and computes the well‑founded model using the described operator. Experimental evaluation on benchmark problems such as graph coloring, set covering, and optimization tasks shows substantial performance gains: aggregate‑based encodings run 3–5 times faster than equivalent aggregate‑free encodings, and memory consumption is reduced due to the compact representation of intermediate results. These results highlight not only the theoretical elegance of the approach but also its practical benefits.
In summary, the paper makes four major contributions: (1) a sound and backward‑compatible generalization of unfounded sets to monotone/antimonotone aggregates; (2) a well‑founded operator that guarantees a unique model and aligns with answer‑set semantics; (3) a complexity analysis establishing polynomial‑time computability for LPAma and coNP‑hardness for unrestricted aggregates; and (4) a working prototype demonstrating that the new semantics yields tangible efficiency improvements in real‑world applications. This work thus bridges a critical gap between expressive aggregate constructs and robust, tractable semantics in answer‑set programming.