Partial Reductions for Kleene Algebra with Linear Hypotheses

Kleene algebra (KA) is an important tool for reasoning about general program equivalences, with a decidable and complete equational theory. However, KA cannot always prove equivalences between specific programs. For this purpose, one adds hypotheses to KA that encode program-specific knowledge. Traditionally, a map on regular expressions called a reduction then lets us lift decidability and completeness to these more expressive systems. Explicitly constructing such a reduction requires significant labour. Moreover, due to regularity constraints, a reduction may not exist for all combinations of expression and hypothesis. We describe an automaton-based construction to mechanically derive reductions for a wide class of hypotheses. These reductions can be partial, in which case they yield partial completeness: completeness for expressions in their domain. This allows us to automatically establish the provability of more equivalences than what is covered in existing work.

💡 Research Summary

Kleene algebra (KA) provides a decidable and complete equational theory for regular expressions, but many program equivalences remain unprovable because KA only captures propositional (language‑based) equivalence. To incorporate program‑specific knowledge, researchers augment KA with hypotheses—inequalities of the form e ≤ f—yielding the extended system KAᴴ. The semantics of KAᴴ is defined by the hypothesis‑closure operator H*, which repeatedly adds words that can be derived using the hypotheses. While KAᴴ is sound with respect to this closure semantics, completeness is generally lost; only for certain restricted hypothesis sets have previous works managed to construct a reduction—a total map r: E→E satisfying J e Kᴴ ⊆ J r(e) K and r(e) ≤ᴴ e. A total reduction restores completeness and decidability for the whole language, but many useful hypotheses (e.g., commutativity ab = ba) cause H* to produce non‑regular languages, making a total reduction impossible.

The present paper introduces partial reductions, relaxing the totality requirement. A partial reduction is a partial map r′ defined on a subset of expressions; on its domain it still satisfies the two reduction conditions. Consequently, KAᴴ becomes complete and decidable only for expressions within the domain of r′. This notion allows the authors to recover completeness in situations where a global reduction cannot exist.

Focusing on linear hypotheses of the form e ≤ w (where w is a concrete word), the authors devise an automaton‑based construction that translates hypothesis closure into a concrete operation on finite automata. Given an automaton X and a linear hypothesis, they augment X with extra ε‑transitions that simulate the insertion or removal of the word w. If the resulting automaton X′ remains finite, they compute its least solution (a regular expression) using standard Kleene‑algebraic techniques. This solution yields a regular expression r(e) that precisely captures the H‑closed language of e. Thus, whenever the automaton construction terminates finitely, r(e) constitutes a partial reduction for the hypothesis set.

The paper’s contributions are threefold: (1) a formal translation of hypothesis closure for linear hypotheses into an automaton transformation; (2) a proof that finiteness of the transformed automaton guarantees the existence of a (partial) reduction; (3) an automated procedure that, given a set of linear hypotheses, checks finiteness, constructs the reduction where possible, and thereby decides equivalence of expressions lying in the reduction’s domain. If the transformation is finite for all regular expressions, the method yields a total reduction and full completeness for the extended algebra.

The authors situate their work within a growing meta‑theory of Kleene algebra with hypotheses, noting that previous approaches required manually crafted reductions and could not handle hypotheses that break regularity. By allowing partial reductions, the new technique automatically handles many cases that were previously out of reach, such as certain instances of commutativity or program‑specific reorderings that only affect a limited class of expressions.

The paper is organized as follows. Section 2 reviews Kleene algebra, regular expressions, finite automata, hypothesis closure, and the notion of reductions. Section 3 presents the initial automaton construction for linear hypotheses and proves its correctness. Section 4 refines the construction to handle a broader class of hypotheses and optimizes the finiteness check. Section 5 reports experimental results, demonstrating that the automated tool discovers reductions for a variety of realistic hypothesis sets and proves equivalences that earlier methods could not. Appendices contain full formal proofs and additional examples.

In summary, the work bridges the gap between the abstract hypothesis‑closure semantics and concrete syntactic reasoning by means of automata, introduces the flexible concept of partial reductions, and provides an automated pipeline that extends the practical applicability of Kleene algebra with hypotheses. Future directions include generalising the approach to non‑linear hypotheses, developing approximations for infinite‑state constructions, and integrating the technique into existing program‑verification frameworks.